Fix docs release builds running out of memory

[cloutiertyler]

Both docs release workflows abort during the Docusaurus build with JavaScript heap out of memory at roughly 2 GiB after running on spacetimedb-linux-small. The release to eddf9f501 consequently updated the branch but did not upload the rebuilt docs or regenerate the LLM index.

Move both docs jobs back to the existing spacetimedb-linux runner and give Node an explicit 8 GiB heap limit. The publish and LLM-generation jobs execute the same full Docusaurus build, so both need the change.

Failure evidence:

• Docs / Publish: V8 heap exhaustion, exit 134; S3 upload skipped.
• Docs / Update llms files: V8 heap exhaustion, exit 134; generation and commit skipped.

Validation:

• Both workflow files parse as YAML; git diff --check passes.
• Full local production docs build with NODE_OPTIONS=--max-old-space-size=8192 passed and generated the static site.

After merging, release docs by preserving the current docs/release tip with a tag, resetting it to master, and pushing. This keeps the workflow fix on the source branch rather than introducing release-only handwritten changes.