[METRICS] Enable synchronous gauge delta temporality and add tests for multi reader path

[niqiangpro-cell]

Fixes #4146

Changes

Delta temporality for synchronous gauge instruments was previously unsupported and silently fell back to cumulative. This PR removes that restriction and verifies correct behaviour on the multi-reader slow path (unreported_metrics_).

• metric_collector.cc
  Removed the guard that rejected kDelta for InstrumentType::kGauge and force-overrode it to kCumulative. Delta temporality for synchronous gauge is now fully supported.

• temporal_metric_storage.cc
  Clarified the fast-path comment: it applies to all instrument types (not just counter/updown-counter/histogram), and documents the window-scoped last-value semantics for synchronous gauge — only values recorded since the previous collection are emitted; stale values are not re-exported.

• sync_metric_storage_gauge_test.cc

  • Added WritableMetricStorageDeltaMultiReaderTestFixture, a dedicated fixture for multi-reader delta tests (prevents cross-instantiation with the existing cumulative fixture).
  • Added LongGaugeLastValueAggregationDeltaTemporalityMultiReader(parameterized, int64_t).
  • Added DoubleGaugeLastValueAggregationDeltaTemporalityMultiReader(parameterized, double).

Both new tests exercise the slow path (collectors.size() > 1) and verify:

1. Each collector independently tracks its own measurement intervals.
2. First interval start_ts equals instrument creation time (spec [Delta temporality uses MeterProvider start time instead of per-instrument creation time for first collection interval #4062]).
3. Subsequent start_ts advances to the previous end_ts.
4. Empty intervals (record → collect → no record → collect) do not invoke the callback — stale gauge values are not re-emitted, for both collector_a and collector_b independently.
5. A late-joining collector (collector_b) receives the merged last-value from all prior intervals with start_ts = instrument creation time.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: niqiangpro-cell / name: Elainen6 (1244ff6)
• ✅ login: niqiangpro-cell / name: niqiangpro-cell (67060a9)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.86%. Comparing base (e8c5f43) to head (a322a7b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4159      +/-   ##
==========================================
+ Coverage   82.82%   82.86%   +0.04%     
==========================================
  Files         406      406              
  Lines       16913    16912       -1     
==========================================
+ Hits        14007    14012       +5     
+ Misses       2906     2900       -6     

Files with missing lines	Coverage Δ
sdk/src/metrics/state/metric_collector.cc	96.67% <ø> (+3.02%)	⬆️
sdk/src/metrics/state/temporal_metric_storage.cc	100.00% <100.00%> (+2.23%)	⬆️

... 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.

[lalitb]

LGTM, with nit comment. Also the PR title should be updated because this PR is not only adding tests. It also enables sync gauge delta temporality by removing the fallback to cumulative.

[niqiangpro-cell]

Hi, @marcalff @ThomsonTan When you have a moment, could you please review this PR?
I believe it's ready, but happy to make any changes if needed.
Thanks a lot for your time!

[niqiangpro-cell]

LGTM, with nit comment. Also the PR title should be updated because this PR is not only adding tests. It also enables sync gauge delta temporality by removing the fallback to cumulative.

Thank you for your review. Do you know if this change could make it into the next release cycle?