Warn when explicit dashboard requests fall back

[peter941221]

Summary

• log a clear scheduler warning when an explicit dashboard request falls back to the missing-bokeh placeholder
• keep the existing HTTP fallback behavior unchanged
• cover missing and incompatible bokeh cases with log assertions in the scheduler HTTP tests

Testing

• python -m py_compile distributed/scheduler.py distributed/http/scheduler/tests/test_missing_bokeh.py
• python -m black --check distributed/scheduler.py distributed/http/scheduler/tests/test_missing_bokeh.py
• python -m pytest distributed/http/scheduler/tests/test_missing_bokeh.py -q # fails locally because this environment is missing the dask dependency required during pytest startup

[github-actions]

Unit Test Results

See test report for an extended history of previous test failures. This is useful for diagnosing flaky tests.

    30 files   -     1      30 suites   - 1   10h 15m 20s ⏱️ - 27m 44s
 4 076 tests +    2   3 967 ✅  -     1    104 💤 ±  0   5 ❌ + 3 
56 545 runs   - 2 586  54 227 ✅  - 2 469  2 305 💤  - 128  13 ❌ +11 

For more details on these failures, see this check.

Results for commit 8a956a1. ± Comparison against base commit d8fabe2.

♻️ This comment has been updated with latest results.

[peter941221]

I believe the PR-local dashboard-warning issues are resolved now.

The remaining red checks in the latest run look unrelated to this PR:

• ubuntu-latest-3.14-default-not ci1: distributed/shuffle/tests/test_shuffle.py::test_handle_null_partitions_2 ([XPASS(strict)] dask/distributed#9184)
• windows-latest-3.10-default-not ci1: workflow/cache path failure
• windows-latest-3.14-default-not ci1: test step passed, artifact upload failed

Run: https://github.com/dask/distributed/actions/runs/26399948001
Summary: https://github.com/dask/distributed/runs/77715380034