Add "timeout" to retryable reasons

[quentin-sommer]

Fixes #779

ℹ️ This modifies the default retry used for all the API calls. Should I instead make a retry object dedicated to the streaming inserts?

[tswast]

I'd like to see a stacktrace of the failure to confirm this is the right way to solve this.

One thing we might be able to do to reproduce is after we implement #897 we can try setting the X-Server-Timeout header to a very small value.

[plamut]

Please just note that the target has been changed from master to main, i.e. the new primary branch.

[quentin-sommer]

I'll try to reproduce it with #897 option 👍 @plamut Thanks for the heads-up

[tswast]

In internal issue 195337762, we did some testing with the X-Server-Timeout header and these were actually reported as "reason": "backendError". Without examples of where "reason": "timeout" appears, I don't think we should accept this PR.

From https://cloud.google.com/bigquery/docs/error-messages, the "timeout" reason happens when "The job timed out.", which to me if anything indicates this should be added to job retry options, but without an example I don't think we should do that either.