Measure the CI cost after the #834 changes

[laughingman7743]

Use case

#834 cannot be closed until the effect of its changes on the AWS bill is measured. The changes so far:

• Run AWS test suites only on ready pull requests and related changes #837 runs no AWS jobs for Draft pull requests, and gates the SQLAlchemy and Spark tests by changed paths.
• Batch SQLAlchemy executemany inserts into multi-row INSERT statements #838 batches SQLAlchemy executemany inserts into multi-row INSERT statements.

Both were merged on 2026-09-26.

Proposed change

Measure the test account cost for two weeks after 2026-09-26 and compare it with the two weeks before.

• Cost Explorer (us-west-2, before tax), per day:
  • S3 PutObject request count and cost
  • Athena SQL cost and Spark DPU-hours
  • Glue Data Catalog requests
• The number of Test workflow runs per day, split by event (pull_request, schedule, workflow_dispatch) and conclusion, and the number of AWS jobs per run. This gives the cost per run and per job.
• Optionally, a sample of CI queries by statement type, to compare with the sample in Reduce the AWS cost of integration tests without reducing real-service coverage #834.

Report the result on #834 and decide whether the remaining sub-issues are still worth their trade-offs.

Validation plan (if implementing)

Read-only Cost Explorer and GitHub API queries; no AWS resources are created. Measure after about 2026-10-10, once two full weeks of data exist.

[laughingman7743]

Measurement window, now that the last cost-related changes of #834 have merged:

• Before: 2026-09-15 to 2026-09-28. The window ends on the day Give the executemany and partition tests their own tables #873 and Define the shared test tables in Python and generate their data #874 merged (Redesign the PyAthena test fixtures: define once, generate data, set up once per job #848's merged steps); the earlier changes (Run AWS test suites only on ready pull requests and related changes #837, Batch SQLAlchemy executemany inserts into multi-row INSERT statements #838, Run pull-request AWS tests on the newest Python version only #863, Run the full test matrix in the Release workflow before publishing #862, Set up the PyAthena test fixtures only in processes that run tests #866, Insert SQLAlchemy compliance fixture rows once for read-only classes #867) also fall in it.
• After: 2026-09-29 to 2026-10-12, with no further cost changes planned. Measure on or after 2026-10-13.

Because some of the earlier changes landed inside the "before" window, it is a mixed baseline. Also compare against 2026-09-01 to 2026-09-14, which predates all of #834.

What to compare, from Cost Explorer with daily granularity, for the test account:

• S3: PutObject requests and cost
• Athena: SQL and Spark DPU
• Glue requests
• Total

Normalize by activity with the Test workflow runs in each window (gh run list --workflow test.yaml): the number of runs and of AWS jobs, split by event (pull_request, schedule, dispatch, release). Report the cost per run as well as the totals.

#834 closes after this measurement.