perf: cache the bundles after we fetched them once

[joanise]

PR Goal?

For a process using convert_prealigned_text_to_offline_html() in the Studio API to create a number of different readalongs, fetching the bundles is pretty slow, and has to be done repeatedly for each readalong created. Running the EveryVoice demo on our cluster, that represented an additional 3 second delay each time I clicked the synthesize button, on top of the actual synthesis time.

The solution I adopted here is simply to save the bundle contents in a file-global variable in make_package.py, so we only ever try to fetch each of the two bundle files once.

Memory cost: the bundles take 600KB together, so after the first fetch I'm reserving .6MB for this cache, which is pretty trivial.

Feedback sought?

sanity checking

Priority?

low

Tests added?

code already covered

Yes

How to test?

Run the everyvoice demo with this branch of Studio vs main, and see synthesis run faster the second and subsequent time.

Confidence?

high

Version change?

Patch, but only when we decide we really want this included in the EV demo...

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
readalongs/text/make_package.py	63% smaller
test/test_api.py	0% smaller

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.22%. Comparing base (301a46a) to head (f592a34).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #260      +/-   ##
==========================================
+ Coverage   88.08%   88.22%   +0.14%     
==========================================
  Files          22       22              
  Lines        1830     1835       +5     
  Branches      330      332       +2     
==========================================
+ Hits         1612     1619       +7     
+ Misses        180      179       -1     
+ Partials       38       37       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[roedoejet]

Nice optimization @joanise - lgtm