only load collections once

[abrookins]

Any time we look up a block from a block document, we load all collections from disk, even if we've already loaded them. It seems unlikely that these will change while the process is running or that we need to support collections changing out from under a running process, so this PR loads them only once. A micro-optimization? Maybe.

Checklist

• This pull request references any related issue by including "closes <link to issue>"
  • If no issue exists and your change is not a small fix, please create an issue first.
• If this pull request adds new functionality, it includes unit tests that cover the changes
• If this pull request removes docs files, it includes redirect settings in mint.json.
• If this pull request adds functions or classes, it includes helpful docstrings.

[codspeed-hq]

CodSpeed Performance Report

Merging #15344 will not alter performance

_{Comparing cache-collections (78d158a) with main (c846de0)}

Summary

✅ 3 untouched benchmarks

[cicdw]

just a style nit, otherwise love it! thank you

[desertaxle]

Would it make sense to put this optimization in the load_collections function directly? That way other consumers share the optimization.

[abrookins]

Would it make sense to put this optimization in the load_collections function directly? That way other consumers share the optimization.

It might! We'd need to cache the output of the function and return that instead. I'll throw up a version like that for you guys to check out. Nothing crazy, but also that object will live in memory, of course.

[abrookins]

Delayed a bit while I add a test suite for the plugins module but should be ready this morning PT.

[abrookins]

I pushed up some commits with the changes we discussed. However, GitHub must be "having a moment" because while the commits appear in the branch, this PR isn't reflecting them. 🙄 A problem for tomorrow.

[abrookins]

@desertaxle @cicdw This is ready again. 👍

[cicdw]

LGTM