fix(retry): stop retrying 404s

Fixes: #19

Spurious 404s are generally rare, and retrying these is more likely to
make things slow than robust. True 404s happen all the damn time and are
often used as quick checks.

README.md

@@ -304,7 +304,7 @@ with, uh, not knocking out people's routers. 🤓
 An object that can be used to tune request retry settings. Retries will only be attempted on the following conditions:
 
 * Request method is NOT `POST` AND
-* Request status is one of: `404`, `408`, `420`, `429`, or any status in the 500-range. OR
+* Request status is one of: `408`, `420`, `429`, or any status in the 500-range. OR
 * Request errored with `ECONNRESET`, `ECONNREFUSED`, `EADDRINUSE`, `ETIMEDOUT`, or the `fetch` error `request-timeout`.
 
 The following are worth noting as explicitly not retried:

index.js

@@ -299,7 +299,6 @@ function remoteFetch (uri, opts) {
         // When hitting an API with rate-limiting features,
         // be sure to set the `retry` settings according to
         // documentation for that.
-        res.status === 404 || // Not Found ("subsequent requests permissible")
         res.status === 408 || // Request Timeout
         res.status === 420 || // Enhance Your Calm (usually Twitter rate-limit)
         res.status === 429 || // Too Many Requests ("standard" rate-limiting)