This is mainly done via sharding, caching & reducing total work:
1. `terser` is turned off when generating test build
2. `nx` distributed cache allows reusing build outputs
3. `playwright` sharding allows running tests in parallel
This can further be improved by using `swc` & `esbuild` during build
**Why?**
The old test suite was a confusing mess, hard to maintain, update, and
add more tests to. It lacked a much needed layer of expressivity &
the reusable functions were all over the place.
Since it used a global `page` (by mistake), it couldn't run in parallel.
Moreover, the global `page` approach caused random flakiness.
All the above reasons led to this OM (Object Model) based approach to tests.
The tests are now much more expressive, reslient, resuable & easier to
maintain. During the rewriting process I also added a couple more
tests (about 10) so this is a net improvement.
Previously, running the tests were also quite slow (15-25s avg). This has
now been improved to (5-8s avg) by running the tests in production.
This means the app now requires to be built before running the tests:
```sh
npm run build:test:web # this is only required once
npm run test:web
```
