Pin silver input tables to fixed Delta versions per report run

[tombonfert]

Summary

Report DataFrames are lazy, so every silver input read through
MeasurementDB._read_table (plain spark.read.table(...) in UC mode, or
spark.read.format("delta").load(...) in path mode) previously pinned no Delta
version. If an input table changed between when a run started and when a given
lazy op materialized, different stages of the same run could read different
snapshots, producing inconsistent results. This is easy to hit when reports run
while upstream ingestion is still writing.

This PR resolves each configured silver table's current Delta version once at
run start and reads every table with versionAsOf, so the whole run observes a
single consistent snapshot. Fixes #87.

Changes

• Add MeasurementDBConfig.pinned_versions (URI to version; empty means "read
  latest", preserving existing behavior).
• Add MeasurementDB.pin_versions(spark), which resolves the current Delta
  version of every configured silver table once via DeltaTable.history(1).
• Update MeasurementDB._read_table to apply .option("versionAsOf", v) for
  both UC and path modes when a version is pinned.
• Call self.db.pin_versions(self.spark) at the start of
  Report.determine_report(), before any solving.
• Exempt debug (in-memory) mode. Non-Delta inputs, views, and unresolvable
  tables are skipped with a warning and keep reading latest, so deployments that
  use them are unaffected.

Testing

• New tests/impulse_query_engine/unit/measurement_db_versioning_test.py:
  snapshot freezing (pin, append rows, assert the read still sees the pinned
  snapshot), debug-mode exemption, and graceful skip plus warning for
  unresolvable tables.
• Full query engine suite (935 passed) and a reporting integration test through
  the complete determine_report path both pass. make lint is clean.

Test Plan

• Unit tests added/updated
• Manual testing completed
• Documentation updated (if applicable)

Checklist

• Code follows project style guidelines
• Self-review completed
• No new linter warnings introduced

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.50%. Comparing base (dfbf0a1) to head (533f425).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #90      +/-   ##
==========================================
+ Coverage   89.37%   89.50%   +0.12%     
==========================================
  Files          62       62              
  Lines        5544     5591      +47     
  Branches      680      686       +6     
==========================================
+ Hits         4955     5004      +49     
  Misses        466      466              
+ Partials      123      121       -2     

Flag	Coverage Δ
query_engine	86.10% <100.00%> (+0.26%)	⬆️
reporting	94.25% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/impulse_query_engine/measurement_db.py	87.85% <100.00%> (+8.58%)	⬆️
src/impulse_reporting/core/report.py	92.08% <100.00%> (+0.02%)	⬆️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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