ci: Drop the unused build-artifact cache

[mpaulosky]

Summary

Item 3 of #144. The build job saved bin/ and obj/ to a cache that no job ever restored, because every test job restores and rebuilds its own project. This removes that step and leaves a comment on the test job explaining why the matrix rebuilds.

How the option was chosen

Each option ran 3 times by workflow_dispatch, one run at a time, on its own branch:

Option	Passed	Wall clock (avg, range)	Runner-minutes (avg)
main as it was	3/3	3.7 min (3.2–4.0)	10.6
Rebuild per project, no cache step (this PR)	3/3	2.9 min (2.8–3.2)	10.2
Share the build output, tests with --no-build	3/3	3.8 min (3.7–3.9)	10.3
One test job running every project	3/3	4.4 min (4.0–4.8)	4.2

• Sharing the build saved nothing. Packing the 164 MB of build output costs about 20s in the build job, and downloading and unpacking it about 9s per test job. The test jobs still need the NuGet cache and a restore, because --no-build skips compiling but the test host and Blazor static web assets still load files from .nuget/packages; its first run failed without them.
• One test job uses 60% fewer runner-minutes, because it pays the per-job setup (checkout, .NET, NuGet cache, restore) once instead of 8 times. It's about 1.5 minutes slower, though, and minutes are free for a public repository.
• Most of the cost in every matrix option is that per-job setup, not the build, which takes about 15s.

Testing

• yamllint, actionlint and zizmor are clean, and scripts/gate.sh passed.
• The option-A branch passed 3 of 3 full CI runs in the experiment.

Refs #144 (atelier-store still to follow).

🤖 Generated with Claude Code

[Copilot]

Copilot review overview

🟢 Approval recommended

The change removes an unrestored cache without altering required build or test behavior.

Review effort: Balanced
Findings: None

What changed in this PR

Removes an unused CI build-output cache while retaining per-project test builds.

Changes:

• Removes caching of bin/ and obj/.
• Documents why test matrix jobs rebuild independently.

File	Description
.github/​workflows/​ci.yml	Simplifies CI caching and documents the tested approach.

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[github-actions]

Test Results Summary

483 tests  ±0   483 ✅ ±0   17s ⏱️ -1s
  9 suites ±0     0 💤 ±0 
  9 files   ±0     0 ❌ ±0 

Results for commit b05effe. ± Comparison against base commit 0ecbdba.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.40%. Comparing base (0ecbdba) to head (b05effe).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #178   +/-   ##
=======================================
  Coverage   85.40%   85.40%           
=======================================
  Files          77       77           
  Lines        1596     1596           
  Branches      150      150           
=======================================
  Hits         1363     1363           
  Misses        189      189           
  Partials       44       44