[eth-json-rpc-middleware] [fetch] Support Retry-After response header

Update the retry logic to look for the `Retry-After` header, and respect the suggested wait time if one is provided.

We should also setting a max wait time, so that the request will fail rather than waiting an inordinate amount of time. A decent max setting might be between 30 seconds to 10 minutes - somewhere in that range. If we wait too long, it becomes increasingly likely that the user may have moved on from the page/operation/etc. that triggered the request (especially considering that we don't yet support cancellation; see MetaMask/core#6873)

For longer wait periods caused by rate limiting, returning an error explaining the delay would be preferable to the request hanging for an hour or longer. That way we can explain the delay to the user directly.

Start in the eth-json-rpc-middleware fetch retry logic and trace how response headers and retry delays are handled. Done means honoring a provided Retry-After value, enforcing a maximum wait, and returning an explanatory error when the delay exceeds that limit.

Written by the indexing model from the issue text.