Improve performance of PR import
What does this MR do?
merge_requests.reload.find makes AR to load ALL Merge Requests for given project.
This results in
PullRequestImporter to have
O(n) complexity and memory pressure
(n - number of already imported PR).
.reload is not needed as the
.find always perform SQL query to find given object.
After that change importing pull requests from https://github.com/gitlabhq/gitlabhq have the very constant time between 0.01-1s for me, instead of 10-40s at the end of the process.
Performance and testing
If cross-browser testing is not required, please remove the relevant item, or mark it as not needed: [-] -->