docs: restore fixes lost in a conflicted merge, and make CI watch main itself

[Chirag6722]

Closes #79

What happened

Merge commit 11f09f7 ("Merge branch 'main' of …", 2026-09-03) resolved README conflicts in favour of a local branch that predated #58, #61 and #63, and reverted all of their README changes. The same merge dropped @LAN-GER's community row for the Isaac Sim-to-Real VLA course (af00e42) — its GIF is still sitting in community/ with nothing pointing at it.

Nothing was force-pushed; every original commit is still an ancestor of main. The content was simply resolved away by hand across five large README files, which leaves no trace in the log. This is an easy mistake to make and not really avoidable by care alone — which is why the second half of this PR matters more than the first.

What this restores

Re-applied onto current main rather than by reverting 11f09f7, so the legitimate newer work in that merge survives — the DLI course link, the Star Arm 102 purchase links, the rebot_arm_b601_dm_isaacsim wiki link are all left as they are.

Restored	Files
Feishu redirect → AliExpress product	4
Sensor/IMU 404 → Sensor_accelerometer	5
shields.io badges re-escaped (literal space, raw emoji)	5
readme.md/#about-power-supply trailing slash	4
RS BOM links → README.md (the real filename)	2
zh reach 650 → 767 mm; reSpeaker row; RS Isaac Sim → completed	1
fr, jp RS Isaac Sim → completed	2
jp RS roadmap table → Japanese (was the French copy again)	1
@LAN-GER's VLA course row, in each language's own wording	5

The Chinese Isaac Sim row keeps the newer DLI Course link while its status returns to ✅ — newer work preserved, the fix restored.

The part that stops it recurring

Neither check could have caught this. Both ran on pull_request only, and these commits were pushed straight to main:

• link-check.yml — pull_request + weekly schedule. It would have caught the broken links on Monday, four days late.
• readme-sync.yml — pull_request only, no schedule at all. It would never have run.

That is a gap in what I set up in #58 and #63, not a process failure by anyone. A guard that only watches pull requests does not protect a repository whose maintainers also push directly. Both workflows now also run on push to main, and readme-sync.yml gets the same weekly schedule link-check.yml has.

Verification

• tools/check_readme_sync.py — 5 disagreements on current main, 0 on this branch.
• Case-sensitive relative-link audit — 2 failures on main, 0 here.
• 9 unit tests pass; actionlint clean on both workflows; the parsed trigger lists confirm push → branches: [main] on each.
• External links: 205 found, 113 checked after .lycheeignore, 0 broken.

Two notes

This is why #78 is red. PR #78 is failing its link check on these same reverted links, not on anything it changes. Merging this turns it green — that red is the check working correctly, on a real regression.

One duplicate to decide on. The VLA course URL now appears twice in README.md: once in the restored community row, once in the roadmap's DLI Course link added afterwards by fa2ef02. Both are legitimate in their own section and they never coexisted before, so I have left both. Say the word if you would rather keep only one.

🤖 Generated with Claude Code