fix(ci): backport every labeled PR in a multi-commit push, not just the head

[Yicong-Huang]

What changes were proposed in this PR?

direct-backport-push.yml resolved exactly one PR per push event — from head_commit.message, falling back to commits/{context.sha}/pulls. When several auto-merges land on main in a single ref update, only the head commit was examined and the release/* labels on every other PR in the same push were silently dropped: no run, no commit status, no failure comment. An audit of 2026-06-01 → 2026-07-04 found 2 of 43 labeled PRs lost this way (#5802, coalesced behind #5812; and #6111, coalesced behind #6112).

Changes to the discover job:

• Enumerate all commits in the push via compareCommits(payload.before...context.sha) (oldest first), falling back to the head commit if the compare fails or before is unavailable.
• Resolve a PR per commit (same two strategies as before: (#N) title suffix, then commits/{sha}/pulls with backoff) and collect green release/* targets per PR.
• Output one matrix entry per (pr_number, merge_sha, target) pair instead of a single PR + target list.

Changes to the push-backports job:

• Matrix is now include: <entries> with max-parallel: 1, so cherry-picks apply in commit order and two legs never race pushes to the same release branch.
• MERGE_SHA/PR_NUMBER come from the matrix entry instead of github.sha/a single discover output; the cherry-pick, commit statuses, and PR comments all annotate the right commit and PR.
• Job legs are named backport #<pr> to <target>; the failure-annotation job-URL matcher was updated accordingly.

Any related issues, documentation, discussions?

Closes #6119

How was this PR tested?

• YAML parses (PyYAML) and all three embedded github-script blocks pass node --check.
• Dry-ran the new discover logic (read-only API calls) against the two historical coalesced pushes: the 2026-07-04 push (5c4a963f7...9cd8acfbb) yields entries for both docs: fix stale core/ directory paths in contribution docs #6111 and docs: update outdated environment requirements in dev guide #6112 with green targets, and the 2026-06-20 push range yields chore(deps): bump tar from 7.5.13 to 7.5.16 in /frontend #5802 while correctly skipping the unlabeled chore(deps): bump torch from 2.8.0 to 2.12.0+cpu in /amber #5809/chore(deps): bump launch-editor from 2.13.2 to 2.14.1 in /frontend #5811/chore(deps): bump js-yaml from 4.1.1 to 4.2.0 in /frontend #5812. The old logic produced only the head PR in each case.
• Note: this fix is not retroactive — chore(deps): bump tar from 7.5.13 to 7.5.16 in /frontend #5802 and docs: fix stale core/ directory paths in contribution docs #6111 still need manual backporting to release/v1.2.

Was this PR authored or co-authored using generative AI tooling?

Generated-by: Claude Code (Claude Fable 5)

[github-actions]

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

• No candidates found from git blame history.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 59.10%. Comparing base (908a61f) to head (55c8068).

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #6120      +/-   ##
============================================
- Coverage     59.12%   59.10%   -0.03%     
+ Complexity     3197     3196       -1     
============================================
  Files          1132     1132              
  Lines         43681    43681              
  Branches       4734     4734              
============================================
- Hits          25828    25818      -10     
- Misses        16419    16426       +7     
- Partials       1434     1437       +3     

Flag	Coverage Δ
access-control-service	70.00% <ø> (ø)
agent-service	44.59% <ø> (ø)
amber	63.07% <ø> (-0.02%)	⬇️
computing-unit-managing-service	0.00% <ø> (ø)
config-service	52.30% <ø> (ø)
file-service	62.81% <ø> (ø)
frontend	51.34% <ø> (-0.04%)	⬇️
notebook-migration-service	78.57% <ø> (ø)
pyamber	91.19% <ø> (ø)
workflow-compiling-service	55.14% <ø> (ø)

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

⚠️ Benchmark changes need a look

🟢 0 better · 🔴 5 worse · ⚪ 10 noise (<±5%) · 0 without baseline

Compared against main 908a61f benchmarked on this same runner, so the delta is largely free of cross-runner hardware noise. The "7d avg" column still reflects the gh-pages dashboard. Treat <±5% as noise unless repeated.

Dashboard · Run

	config	throughput	MB/s	latency	max Δ latest / 7d
🔴	bs=10 sw=10 sl=64	389	0.238	24,994/34,400/34,400 us	🔴 +8.6% / 🔴 +127.2%
🔴	bs=100 sw=10 sl=64	807	0.492	123,371/152,583/152,583 us	🔴 +8.7% / 🔴 +41.4%
⚪	bs=1000 sw=10 sl=64	927	0.566	1,076,850/1,148,561/1,148,561 us	⚪ within ±5% / 🔴 +11.7%

Baseline details

Latest main 908a61f from same runner

config	metric	PR	latest main	7d avg	Δ latest	Δ 7d
bs=10 sw=10 sl=64	throughput	389 tuples/sec	423 tuples/sec	776.36 tuples/sec	-8.0%	-49.9%
bs=10 sw=10 sl=64	MB/s	0.238 MB/s	0.258 MB/s	0.474 MB/s	-7.8%	-49.8%
bs=10 sw=10 sl=64	p50	24,994 us	23,017 us	12,636 us	+8.6%	+97.8%
bs=10 sw=10 sl=64	p95	34,400 us	34,371 us	15,143 us	+0.1%	+127.2%
bs=10 sw=10 sl=64	p99	34,400 us	34,371 us	18,954 us	+0.1%	+81.5%
bs=100 sw=10 sl=64	throughput	807 tuples/sec	824 tuples/sec	985.33 tuples/sec	-2.1%	-18.1%
bs=100 sw=10 sl=64	MB/s	0.492 MB/s	0.503 MB/s	0.601 MB/s	-2.2%	-18.2%
bs=100 sw=10 sl=64	p50	123,371 us	119,429 us	101,671 us	+3.3%	+21.3%
bs=100 sw=10 sl=64	p95	152,583 us	140,373 us	107,939 us	+8.7%	+41.4%
bs=100 sw=10 sl=64	p99	152,583 us	140,373 us	113,798 us	+8.7%	+34.1%
bs=1000 sw=10 sl=64	throughput	927 tuples/sec	922 tuples/sec	1,016 tuples/sec	+0.5%	-8.8%
bs=1000 sw=10 sl=64	MB/s	0.566 MB/s	0.563 MB/s	0.62 MB/s	+0.5%	-8.8%
bs=1000 sw=10 sl=64	p50	1,076,850 us	1,081,936 us	989,693 us	-0.5%	+8.8%
bs=1000 sw=10 sl=64	p95	1,148,561 us	1,144,141 us	1,028,327 us	+0.4%	+11.7%
bs=1000 sw=10 sl=64	p99	1,148,561 us	1,144,141 us	1,059,969 us	+0.4%	+8.4%

Raw CSV

config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,513.93,200,128000,389,0.238,24993.53,34400.29,34400.29
1,100,10,64,20,2479.12,2000,1280000,807,0.492,123371.13,152582.92,152582.92
2,1000,10,64,20,21563.51,20000,12800000,927,0.566,1076849.89,1148560.70,1148560.70

[xuang7]

LGTM!