Optimize registry fetch latency

[jmrossy]

Capturing thoughts from discussion here: https://discord.com/channels/935678348330434570/1238863226419023892

There are many options for ways we can optimize this:

1. Assemble all chain metadata and/or address files into one file during registry build
2. Add backend caching to warp ui + explorer apps
3. Pre-fetch data at build time
4. Refactor apps to allow lazy-loading of data

[jmrossy]

I like option 1 the best because it makes the improved latency available to all registry users, not just those that have custom caching / build-time logic implemented. It's also the easiest to do.

[jmrossy]

Snippet from more discord discussion: https://discord.com/channels/935678348330434570/1243986495916736632/1246957837805162669

...having thought through it more, all the options have downsides. Build-time chain data assembly would require contributors to have the JS tooling (e.g. yarn). PR CI-time requires a github token with write perms, which would complicate things for registry forkers. Ditto for a post-PR-merge CI job.
I'm still open to Yorke's suggestion of downloading the full repo but I think the size will scale up fast as more chain + token logos are added. And I suspect the decompression step will have pitfalls of its own.

[jmrossy]

In the end I was able to make CI-based updates work well enough.