ci(jest): Fix importing of BALANCE_RESULTS_PATH (#96499)

This `require()` call wasn't working since we added
`--experimental-transform-types` in
#93303

We can see our balancer_strategy logging stopped logging `by_path` as
often, instead showing 'other':
<img width="530" height="196" alt="SCR-20250726-mgai"
src="https://github.com/user-attachments/assets/b449b7a7-f750-47c1-9b4c-1e602fe55fbd"
/>


And this impacted at test worker runtimes too:
<img width="531" height="203" alt="SCR-20250726-mfyl"
src="https://github.com/user-attachments/assets/98e4b294-1b16-4b3e-b487-c443132e7d3e"
/>


Without the balancer stats js tests got more than a minute slower,
because the shards are unbalanced and slowest one is taking longer now.
What we learned though is that the jest-balance.json sharing is pretty
good and keeps all the shards at about the same runtime!


I made a new dashboard to see all the CI worker runtimes for
frontend.yaml:
https://sentry.sentry.io/dashboard/141271/?project=5899451&statsPeriod=30d.
The critical path is usually jest, but eslint is taking almost the same
amount of time too running in parallel. If we speedup jest then we'll
save wall-time in CI.
There's also a dashboard to see the slowest test suites and test cases.
It's got the balancer_strategy chart added too:
https://sentry.sentry.io/dashboard/141273/?project=4857230&statsPeriod=30d

Author: ryan953

jest.config.ts

@@ -219,7 +219,7 @@ if (
     'jest-balance.json'
   );
   try {
-    balance = require(BALANCE_RESULTS_PATH);
+    balance = (await import(BALANCE_RESULTS_PATH, {with: {type: 'json'}})).default;
   } catch (err) {
     // Just ignore if balance results doesn't exist
   }