[Project516]

What this changes

The weekly upstream watch opened #8 and #10 for versions that need no action. This records them as reviewed, so the next run closes both issues and only reopens them for something newer.

• PhotonVision: same-season prereleases are skipped. v2026.7 is PhotonVision's April Fools release (prerelease, published 2026-04-01, nine days before v2026.3.4). version_key ranked it above v2026.3.4 because 7 > 3.
• PhotonVision next-season tags are compared with a new PV_NEXT_REVIEWED=v2027.0.0-alpha-2 in scripts/ci/upstream-pins.env. alpha-2 is the robot's PhotonLib already. The Jetson's move to 2027 is Document WPILib 2027 alpha-7 migration limits #1 and Experimental alpha-7 CUDA and PhotonVision migration #2.
• allwpilib tags are compared with a new ALLWPILIB_REVIEWED=v2026.2.2. 04-build-allwpilib.sh must match the PhotonVision fork's wpilibVersion, which is 2026.2.1 at d8c9e8e, so bumping the tag alone would split the detector from PhotonVision's own WPILib. v2026.2.1...v2026.2.2 changes no wpiutil or apriltag files. In wpimath it only touches PoseEstimator, PoseEstimator3d and SlewRateLimiter, which the detector does not use.
• Both issue bodies now say which pin to move once a new tag is read, as the aos one already does.

Closes #8
Closes #10

Rebased onto main at ab6e086 on 2026-10-12. No conflicts. #14 also edits upstream_watch.py (the L4T check) and merges with this cleanly.

Checks

• python3 -m py_compile scripts/ci/upstream_watch.py and scripts/ci/lint.sh, both clean.
• Called report_photonvision() and report_allwpilib() against the live GitHub API: both return nothing with the new pins. With the pins set back one step (v2027.0.0-alpha-1, v2026.2.1), they report v2027.0.0-alpha-2 and v2026.2.2 with the new hint, and v2026.7 no longer appears.
• The workflow itself was not run, since it edits issues on whatever repo it runs in.

[Project516]

@allengregoryiv ready for a review when you have a moment. Rebased onto main at ab6e086 today with no conflicts. It closes #8 and #10.