[Roy-Kid]

Problem

The Deploy docs job pushes the built site to gh-pages with mike deploy -p. gh-pages is a single shared ref, so two master merges landing close together deploy in parallel and the slower job's push is rejected:

error: failed to push branch gh-pages to origin:
 ! [rejected]  gh-pages -> gh-pages (fetch first)

That is what broke run 35215109985 — #47 and #46 merged 36 seconds apart, #47's deploy won the race, and #46's docs fix never reached the site. Everything else in the run (pytest on Py3.9/3.10/3.11 + TF 2.15, both docker builds) was green.

Change

Waiting for a PR's docs deploy to finish before merging the next one avoids the race, so this does not serialize deploys automatically. It only opens the repair path for when the race happens anyway.

The workflow already listed workflow_dispatch under on:, but deploy-docs was gated on github.event_name == 'push' — so "Run workflow" rebuilt and re-tested without ever redeploying the docs. The gate now accepts workflow_dispatch as well, with the Teoroo-CMC/PiNN and master/v*-ref conditions unchanged. One line plus a comment explaining the race; no new file and no duplicated deploy steps.

Note that a manual run still goes through needs: pytest first, so a replay costs a full test matrix (~10 min). That seemed better than duplicating the checkout/install/mike deploy block into a second workflow just to skip the tests.

Test plan

• YAML parses; on: keeps push / pull_request / workflow_dispatch, and deploy-docs keeps its steps and needs: pytest
• gh-pages is currently back in sync (re-ran the failed job — the site is at b166b86), so the condition can be tested from a clean state
• After merge, "Run workflow" on master redeploys the docs and gh-pages advances
• A workflow_dispatch run from a non-master branch still skips deploy-docs