Add retry logic for github errors to improve reliability.

[calebbrown]

1. Adds a RetryAfter function for processing responses with Retry-After header (will move to the retry library in the future)
2. Adds a SecondaryRateLimit function for detecting rate limiting from GitHub and pausing immediately
3. Adds a ServerError400 function for detecting odd server errors from GitHub
4. Adds a ServerError function for detecting any 5xx server error and retrying (except for issues/comments which knowingly fails on lots for large numbers of comments).

All of these are well tested to ensure the retry behavior works correctly.

[calebbrown]

For reference: I have a job still running at nearly 72 hours non-stop on 1 token with this error handling, so it appears to be sufficient.