Clean up schema for table milestones

Everyone can contribute. Help move this issue forward while earning points, leveling up and collecting rewards.

• Close this issue

From https://gitlab.com/gitlab-com/infrastructure/issues/1709 (private to GitLab employees):

• Missing constraint: iid NOT NULL
• Missing constraint: state NOT NULL
• Extra default: Specifying "default NULL" isn't necessary- NULL is the default when there is no default
• Missing FK: project_id -> projects (40 entries violate this rule)
• Extra index: btree index on title- is it actually useful?
  • If MySQL doesn't use it we can get rid of this since we use trigram/GIN indexes for searching.
• Duplicate data: state would be more efficient as an enum or integer (or smaller), save 649 kB
• Extra index: due_date index does not look very used, perhaps not necessary?
  • Probably because milestone due dates are not commonly used on GitLab.com, meaning we still need to keep the index for self hosted indexes.
  • First figure out how frequently these are used, then see if removing them slows things down or not
• due_date is frequently NULL, do NULLs need to be in the index? Perhaps use partial index
  • NULLs should not be indexed since you can not filter by due_date with NULL values.