Fix GHA stuff

[penelopeysm]

This PR makes a few changes to GitHub Actions workflows:

1. Separates the preview workflow from the remove-preview workflow.

   Whenever a PR was merged to master, the remove-preview workflow would fire. However, the publish workflow on master would also fire. The changes in Add concurrency key to publish/preview workflows #527 meant that the publish workflow would cancel the remove-preview workflow. The result is that we have a buildup of PR previews.

   This fixes it by making the remove-preview workflow a separate one entirely, so that its running is unaffected by other workflows.

2. Modifies the version check workflow to not use continue-on-error: true. This causes the workflow to always report a green tick, even if the step did have an error. Instead I use always() on the subsequent steps.

3. Slightly improves the version checking script.

[penelopeysm]

The effects of (2) are already visible 😄 it is failing because it can't resolve an env with Turing v0.35.2, as expected, see #559 (comment).

Unfortunately, (1) is not visible because it uses the pull_request_target event rather than pull_request. That means that GitHub runs the workflow on the base branch (i.e. master) rather than the workflow in the PR (i.e. the modified one).

[github-actions]

Preview the changes: https://turinglang.org/docs/pr-previews/563
Please avoid using the search feature and navigation bar in PR previews!

[shravanngoswamii]

Looks good, thank you, @penelopeysm!

Fixing these issues in GHA is really great. I was aware of the PR-preview removal issue, but we decided to ignore it when merging this repo into Quarto. At that time, we were mainly focused on addressing internal documentation problems, so the GHA workflows were written to just work, but were not fully optimized or cleaned up. It's really helpful to have this addressed now—thanks for your efforts!

[penelopeysm]

Thanks @shravanngoswamii! To be honest, I'm pretty sure that this is still not quite right. There's probably some weird behaviour that I haven't figured out. But with GHA I always think it's fine to merge something imperfect since it can't really cause any bugs in the actual code 😅

[shravanngoswamii]

To be honest, I'm pretty sure that this is still not quite right. There's probably some weird behaviour that I haven't figured out. But with GHA I always think it's fine to merge something imperfect since it can't really cause any bugs in the actual code 😅

Let's see, we can always change it if it creates any issue, experiments takes the product near perfection😅!