Rework the metric interface and collection to use explicit accumulators

[mdekstrand]

This overhauls the metric interface and the metric collection code to use explicit accumulator objects, so different metrics can have different accumulation behavior, and allowing things like Gini to be computed with less memory consumption.

It also simplifies and deprecates RunAnalysis in favor of directly using MeasurementCollector.

Several of the old MeasurementCollector tests are removed because the measurement collector is responsible for much less logic now.

Closes #866, #968.

[codecov]

Codecov Report

❌ Patch coverage is 92.43028% with 19 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.49%. Comparing base (ab4c43e) to head (8d71a1b).
⚠️ Report is 23 commits behind head on main.

Files with missing lines	Patch %	Lines
src/lenskit/data/accum/_object.py	0.00%	11 Missing ⚠️
src/lenskit/metrics/_collect.py	96.20%	3 Missing ⚠️
src/lenskit/data/accum/_value.py	93.33%	2 Missing ⚠️
src/lenskit/metrics/bulk.py	87.50%	2 Missing ⚠️
src/lenskit/metrics/ranking/_gini.py	96.29%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1063      +/-   ##
==========================================
- Coverage   89.54%   89.49%   -0.05%     
==========================================
  Files         229      233       +4     
  Lines       15870    15890      +20     
==========================================
+ Hits        14210    14220      +10     
- Misses       1660     1670      +10     

☔ 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.