As BitBucket importer tried to fetch all Pull Requests upfront, before
importing it, the memory consumption got very high. So high, that bigger
projects could never be imported fully, as they had too many Pull
Requests.

This change should reduce the amount of occupied memory, as the Pull
Requests are now fetched and imported in a batch of 100. As it used to
be done before hand, for creating pull request branches.

relates to https://gitlab.com/gitlab-org/gitlab/-/issues/23586