Read GeoTIFF and TIFF through the nd pipeline

[robinskil]

beacon-arrow-tiff was the last nd format still on the v1 broadcast: it built the pixel
cross-product per column inside the opener. It now runs the same spine as netCDF, HDF5 and Zarr.

What changed

The spine. The opener emits beacon.nd-encoded batches through
any_dataset_as_encoded_stream and adapts them in the encoded struct domain.
TiffFormat::create_physical_plan sets the file source schema to encoded_schema(file_schema)
and returns NdBroadcastExec(NdSourceExec(DataSourceExec)).

A raster is a grid over y (image rows) and x (image columns): bands on (y, x), geo.lat on
y, geo.lon on x, and the TIFF tags as rank-0 scalars. The cross-product is therefore built
once, at the top of the plan, and the two nd optimizer rules now reach a raster:

• WHERE "geo.lat" > 40 sinks into an NdFilterExec and selects the grid before it materializes.
• SELECT "geo.lat" * 2 sinks into an NdProjectionExec and evaluates over the 380-value
  latitude axis instead of all 489060 pixels.

COUNT(*) keeps the v1 zero-column driver path, and now also drives with the predicate
columns. The old any_dataset_as_row_size path ignored the predicate, so a COUNT(*) with a
WHERE over-counted.

Dimension parity. read_tiff('raster.tif', ['y']) returns one row per image row, matching
read_netcdf / read_hdf5 / read_zarr. The same list works as
OPTIONS (read_dimensions 'y') on CREATE EXTERNAL TABLE ... STORED AS TIFF, and
resolve_read_dimensions now runs in both schema inference and the opener. read_tiff_schema
inherits the argument automatically.

Behaviour change worth reviewing

The nd encoding does not carry the target grid; decode_nd_record_batch_row re-infers it from the
projected columns. So a projection can change the row count:

query	before	after
SELECT count(*) FROM tiff	489060	489060
SELECT count("geo.lat"), count("band.0") FROM tiff	489060	489060
SELECT count("geo.lat") FROM tiff	489060	380

The last row is the change: with the full-rank band projected away, the grid narrows to the axes
the surviving columns carry. Same-rank columns also leave the row order unspecified, because
infer_target takes the axis order from the widest column.

This is a property of the shared encoding, not of the TIFF reader. netCDF, HDF5 and Zarr already
behave this way, and the zarr test attribute_is_single_distinct_value_across_grid documents it.
Fixing it means carrying the target grid through encode_nd_record_batch, which changes all four
formats, so it is out of scope here.

Tests

• beacon-arrow-tiff: 23 pass. New coverage for the plan shape, projection pushdown, filter
  pushdown, COUNT(*), scalar broadcast, and read_dimensions (schema and row count).
• beacon-core/tests/read_functions.rs: two end-to-end tests through a real runtime, beside the
  HDF5 ones.
• cargo test --workspace --lib --bins --tests passes.
• cargo clippy --workspace --lib --bins --tests: no new warnings; the crate drops from 7 to 5,
  all pre-existing.

Docs (formats/geotiff.md, sql/table-functions.md) and the CHANGELOG record the new argument
and the grid semantics.

[codecov]

Codecov Report

❌ Patch coverage is 85.34572% with 142 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.06%. Comparing base (0ac879e) to head (e067706).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...-db/beacon-datafusion-ext/src/fast_object/prune.rs	79.32%	43 Missing ⚠️
...ile-formats/beacon-arrow-odv/src/datafusion/mod.rs	0.00%	24 Missing ⚠️
...formats/beacon-arrow-tiff/src/datafusion/source.rs	71.79%	22 Missing ⚠️
...beacon-arrow-tiff/src/datafusion/table_function.rs	59.37%	13 Missing ⚠️
...-db/beacon-datafusion-ext/src/fast_object/table.rs	94.15%	9 Missing ⚠️
...le-formats/beacon-arrow-tiff/src/datafusion/mod.rs	97.34%	7 Missing ⚠️
...formats/beacon-arrow-tiff/src/datafusion/reader.rs	71.42%	4 Missing ⚠️
...eacon-arrow-atlas/src/datafusion/table_function.rs	0.00%	3 Missing ⚠️
.../beacon-arrow-bbf/src/datafusion/table_function.rs	0.00%	3 Missing ⚠️
...-arrow-geoparquet/src/datafusion/table_function.rs	0.00%	3 Missing ⚠️
... and 6 more

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #363      +/-   ##
==========================================
+ Coverage   80.87%   81.06%   +0.19%     
==========================================
  Files         371      370       -1     
  Lines       57527    57811     +284     
==========================================
+ Hits        46525    46865     +340     
+ Misses      11002    10946      -56     

Files with missing lines	Coverage Δ
beacon-db/beacon-common/src/file_descriptors.rs	89.47% <100.00%> (+0.58%)	⬆️
beacon-db/beacon-common/src/super_typing.rs	69.20% <ø> (-3.05%)	⬇️
beacon-db/beacon-core/src/runtime_builder.rs	88.46% <100.00%> (+0.08%)	⬆️
beacon-db/beacon-core/src/statement_plan/authz.rs	93.41% <100.00%> (+1.84%)	⬆️
...acon-db/beacon-datafusion-ext/src/ordered_union.rs	76.11% <ø> (ø)
...acon-db/beacon-datafusion-ext/src/unique_values.rs	70.50% <ø> (ø)
...e-formats/beacon-arrow-atlas/src/datafusion/mod.rs	92.20% <100.00%> (+0.45%)	⬆️
.../beacon-arrow-csv/src/datafusion/table_function.rs	87.50% <100.00%> (ø)
...eacon-file-formats/beacon-arrow-hdf5/src/format.rs	93.23% <100.00%> (ø)
...le-formats/beacon-arrow-hdf5/src/table_function.rs	63.63% <100.00%> (ø)
... and 21 more

... and 4 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.