schema changes: these changes will do a join on the repo table so that the repository ids are populated with the id of the repository that has the same name. If a repository id cannot be correlated (due to a rename that has already occurred), those uploads are removed.
rename lsif_uploads.repository to repository_name_at_upload
add new field lsif_uploads.repository_id
add new field lsif_commits.repository_id
drop lsif_commits.repository
remove all indexes/constraints from repository and add symmetric ones to repository_id (both tables)
lsif-server api changes:
accept repository_id and (optionally) repository_name instead of repository for all query endpoints
accept repository_id and (required, see see Slack thread) repository_name instead of repository for the upload endpoint
lsif-server internal changes: use repository_id everywhere where repository was used before (except for gitserver interfacing, which needs to use the repository name value, if supplied to the API)
lsif-server tests: this is where most of the diff is - but they're all trivial differences
graphql api changes: remove InputRepoName, as we no longer have this data and recent changes to the uploads graphql api has made it an unnecessary datapoint (you must specify a repo in order to list uploads in the first place)
