Faster translations

[aatkin]

Internal performance update to how translations are fetched.

Motivation

Email template transformations take quite a bit of time when large-ish number of emails are produced. For example application.event/decided creates 1 email for each workflow handler:

event-to-emails, disabled cache

handler count	mean	lower-q 2.5%	upper-q 97.5%
1	4,780888 ms	4,097078 ms	6,170799 ms
10	52,443133 ms	49,652797 ms	55,356244 ms
50	223,241818 ms	203,445667 ms	249,808264 ms
100	441,938544 ms	413,564604 ms	478,001015 ms
200	834,458932 ms	804,961543 ms	914,037245 ms

(using 2019 Macbook Pro with 2,3 GHz 8-Core Intel Core i9)

Caching dictionary for tempura function reduces mean time by almost 100x. For same event, but using cached translation function:

event-to-emails, cache

handler count	mean	lower-q 2.5%	upper-q 97.5%	cache size
1	40,996545 µs	39,542207 µs	43,635827 µs	882,5 KiB
10	397,217641 µs	367,224108 µs	435,834954 µs	882,5 KiB
50	2,158321 ms	1,904828 ms	2,440271 ms	882,5 KiB
100	4,206048 ms	3,757476 ms	4,956851 ms	882,5 KiB
200	8,656116 ms	7,626819 ms	10,665969 ms	882,5 KiB

Checklist for author

Remove items that aren't applicable, check items that are done.

Reviewability

• [ ] Link to issue (no issue number)
• [x] Consider adding screenshots for ease of review (clj-async-profiler)

Documentation

• [ ] Update changelog if necessary

Testing

• [x] Complex logic is unit tested
• [x] Valuable features are integration / browser / acceptance tested automatically

Follow-up

• [ ] New tasks are created for pending or remaining tasks