Add telemetry context propagation across search and load

[charles-turner-bot]

Summary

This adds a small telemetry-lineage mechanism so catalog usage can be correlated across:

• search()
• selection of a result/catalog entry
• to_xarray() materialisation

The implementation is deliberately lightweight:

• adds a TelemetryContext dataclass anchored on a stable store_id
• propagates that context through IcechunkCatalog.search() and downstream data-source creation
• emits optional catalog.search, catalog.to_xarray.start, and catalog.to_xarray.end events via a callback hook
• attaches lineage attrs to materialised xarray datasets for debugging/inspection

Why this shape

This avoids relying on Python object identity while still giving us a clean way to trace flows where the same backing icechunk store survives pre/post-search as different Python objects.

I kept it as context propagation + optional emitter hook rather than an accessor-only design, since the main problem seems to be lineage survival across object boundaries.

Testing

• pixi run -e test-py313 lint
• pixi run -e default pytest -q

Closes #27.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.70%. Comparing base (15a9c58) to head (fcbc9ad).
⚠️ Report is 30 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #28      +/-   ##
==========================================
+ Coverage   88.37%   89.70%   +1.33%     
==========================================
  Files           8        9       +1     
  Lines         456      515      +59     
==========================================
+ Hits          403      462      +59     
  Misses         53       53              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.