fix(enterprise): improve user creation process

[sranka]

Closes #5840

What was the problem?
Sometimes, when a multi-meta-node enterprise cluster has many users, the chronograf user creation process fails on user not found, the user is however created. This happens because of the eventual consistency of the meta-node cluster, the follower nodes (that are asked to get the user) might not be in sync yet with the leader node (that is always asked to create the user).

What was the solution?

• remember the last redirect to the master meta node (performed when the user is created) and then stick the communication (getting the user, assigning permissions/roles) to the master node in order to avoid stale reads in an eventually consistent meta-nodes cluster
• (wait for eventual consistency) wait for the user creation to be consistent across all meta nodes (at most for 2 seconds), re-check every 50ms
• (optimize) do not update empty user permissions during user creation

• CHANGELOG.md updated with a link to the PR (not the Issue)
• Well-formatted commit messages
• Rebased/mergeable
• Tests pass

[vlastahajek]

LGTM