Consider adding retry and timeout wrappers to fetch calls

[jmrossy]

The GithubRegistry makes fetches via the native fetch API here: https://github.com/hyperlane-xyz/hyperlane-registry/blob/sprint27/src/registry/GithubRegistry.ts#L211

To compensate for flaky connections, it may make sense to add some retry logic for failures.
The downside is that if the problem is legitimate (e.g. bad URL), it would burn through an IP's quota faster.

Some timeout handling around the fetch would also be a good idea.

Note, #29 would help mitigate connectivity issues entirely so I'd recommend we start there.