Work with both DASCore branches without importing removed classes

[d-chambers]

Description

DASCore's development branch collapses MemorySpool/DirectorySpool/FileSpool into a single dc.Spool and deletes the dascore.clients package, so from dascore.clients.dirspool import DirectorySpool fails at import. That takes down widget discovery entirely — every DerZug entry point fails to load:

ERROR orangecanvas.registry.discovery: An exception occurred while loading entry point 'Spool'
  File "src/derzug/widgets/spool.py", line 31, in <module>
    from dascore.clients.dirspool import DirectorySpool
ModuleNotFoundError: No module named 'dascore.clients.dirspool'

Four more couplings were hiding behind that first ImportError. This PR fixes all five by reaching for public APIs instead of version-specific classes and private helpers, with no branching on a DASCore version anywhere in the code.

Coupling	Fix
dascore.clients.dirspool / filespool imports — package deleted	New derzug/utils/dascore_compat.py answers spool-kind questions from public accessors (indexer, spool_path)
get_contents() renamed path → source_path; row ordering and identity tokens read it by name	contents_path_column() resolves whichever name is present
Spool.select now validates names against a spool's attrs/coords, so the widgets' distance_min dropdown entries raise	selectable_contents_keys() offers the coordinate itself (distance) plus real attributes
slope_filter(notch=...) removed — it was a deprecated alias for invert	Both checkboxes fold into invert=
dascore.viz.waterfall._get_scale — private, and imports matplotlib eagerly	Calls the public dascore.utils.misc.tukey_fence, which is exactly what that call path computed

Also swapped dascore.core.coords.get_coord for the top-level dc.get_coord.

derzug.utils.dascore_compat is deliberately the single place that knows about the two layouts, so it is the one file to delete when support for the older one is dropped.

User-visible changes

• Select keys. The Spool and Select widgets no longer offer the flattened distance_min/time_max columns or source-file bookkeeping columns; select on the coordinate itself (distance with a (1000, None) range). Auxiliary coordinates such as latitude are now offered, which they never were before. Workflows saved with an older key keep filtering on it.
• The Filter widget still shows both a Notch and an Invert checkbox. They now do the same thing — as they effectively already did, since notch=True forced invert — but collapsing them is a UI change I kept out of a compatibility PR. Happy to fold it in if wanted.

Verification

The suite was run against three DASCore versions:

DASCore	Result
dev @ dddf0f73 (0.1.21.dev146)	2135 passed, 44 skipped
master @ 64f0c8ed	2135 passed, 44 skipped
released 0.1.16	compat surface only — see note

The 0.1.16 environment cannot run the Qt suite (unrelated broken chardet/Orange install), so it was exercised with a direct script over the Qt-free compat surface and the node-layer spool functions.

All 26 widget entry points load against dev, which is what the original traceback was about.

Checklist

I have (if applicable):

• referenced the GitHub issue this PR closes.
• documented new behavior with docstrings and/or a doc page.
• included tests. See guidelines.
• run prek run --all-files and pytest tests.
• for hot-path changes, run python scripts/bench_compare.py --baseline main, pasted the table below, and added the benchmark label. See benchmarking.

Note on prek run --all-files: it also reformats tests/test_views/test_orange_qt.py, which it does on a clean main too — the pinned ruff (v0.4.8) is older than whatever last formatted that file. Left untouched here; it wants its own fix.

Benchmarks

python scripts/bench_compare.py --baseline main --repeat 2 (control -5.8%, within tolerance):

benchmark                                                baseline       head    median       min status
-------------------------------------------------------------------------------------------------------
...rks.py::TestGraphBenchmarks::test_validate_uncached    325.0us    336.1us    +3.4%    -2.5%   --  
...chmarks.py::TestExecutionBenchmarks::test_run_chain    312.8us    321.6us    +2.8%    -1.5%   --  
...chmarks.py::TestExecutionBenchmarks::test_map_chain     1.47ms     1.51ms    +2.5%    -3.0%   --  
...ks.py::TestAnnotationMaskBenchmarks::test_span_mask    396.2us    404.6us    +2.1%    +3.1% noisy   (baseline rel stdev 12%)
...ks.py::TestExecutionBenchmarks::test_get_provenance     70.3us     71.1us    +1.2%    -4.0% noisy   (baseline rel stdev 13%)
...rks.py::TestGraphBenchmarks::test_serialize_to_dict    101.7us    102.3us    +0.6%    -0.0%   --  
...mplingBenchmarks::test_percentile_levels_subsampled    929.8us    934.7us    +0.5%    -2.3%   --  
...chmarks.py::TestTaskBenchmarks::test_port_spec_cold    156.8us    156.3us    -0.4%    -0.6%   --  
...::TestExecutionBenchmarks::test_build_runtime_specs    862.5us    854.8us    -0.9%    -4.1%   --  
...marks.py::TestExecutionBenchmarks::test_run_diamond    205.8us    203.9us    -0.9%    -2.5%   --  
...arks.py::TestGraphBenchmarks::test_topological_sort    265.2us    262.2us    -1.1%    -3.5%   --  
...s.py::TestTaskBenchmarks::test_extract_return_names     1.21ms     1.19ms    -1.4%    +1.4%   --  
...s.py::TestTaskBenchmarks::test_callable_source_hash     1.10ms     1.08ms    -1.6%    +0.3% noisy   (baseline rel stdev 16%)
...rks.py::TestSamplingBenchmarks::test_strided_sample    156.2us    153.6us    -1.7%    -2.8%   --  
...arks.py::TestGraphBenchmarks::test_build_chain_pipe    740.4us    725.8us    -2.0%    -2.2%   --  
...py::TestGraphBenchmarks::test_deserialize_from_dict    123.7us    120.3us    -2.7%    -4.2% noisy   (baseline rel stdev 13%)
...nnotationMaskBenchmarks::test_box_mask_object_dtype    11.15ms    10.80ms    -3.1%    -2.7%   --  
...marks.py::TestTaskBenchmarks::test_task_fingerprint     5.96ms     5.75ms    -3.5%    -1.4% noisy   (baseline rel stdev 34%)
...marks.py::TestTaskBenchmarks::test_port_spec_cached    938.2us    905.0us    -3.5%    -3.4%   --  
...stTaskBenchmarks::test_function_task_class_creation    355.8us    339.7us    -4.5%    +0.8%   --  
...s.py::TestImportBenchmarks::test_import_spool_utils     5.77ms     5.45ms    -5.5%    -9.1%   --  
...::TestGraphBenchmarks::test_ensure_validated_cached    743.2us    702.1us    -5.5%    -5.0%   --  
...s.py::TestExecutionBenchmarks::test_stream_pipeline    105.9us     99.6us    -5.9%    -3.0% noisy   (baseline rel stdev 14%)
...TestAnnotationMaskBenchmarks::test_box_mask_numeric     3.02ms     2.82ms    -6.5%    -2.7% noisy   (baseline rel stdev 14%)
...arks.py::TestImportBenchmarks::test_import_workflow    12.32ms     9.92ms   -19.5%   -21.0% noisy   (baseline rel stdev 31%)

control: -5.8% (ok)

threshold 20%: 0 slower, 0 faster, 25 compared

Nothing regresses past the 20% threshold; the largest slowdown is +3.4% on a benchmark this change does not touch. The one double-digit row (test_import_workflow, -19.5%) is an improvement and flagged noisy (baseline rel stdev 31%).