pull request commits collection could be more efficient

We might be able to squeeze a lot more performance from our api calls in PR commits collection

Sonnet 4.6 has identified that we are constructing one graphql call per pull request:

https://github.com/chaoss/CollectOSS/blob/96adf3a4d68725db21622673ee6613693c0f5ace/collectoss/tasks/github/pull_requests/commits_model/core.py#L49-L56

Some of these may return very few commits

Other places in the code already batch multiple queries into one graphQL call:

https://github.com/chaoss/CollectOSS/blob/96adf3a4d68725db21622673ee6613693c0f5ace/collectoss/tasks/frontend.py#L207-L212

We should do the same in pull request commits to help make things faster (less waiting for network calls)

(note: this issue should not be confused with the similar issue #420, which describes the same thing for a different part of secondary collection)