feat: nlu-server can compute dataset issues to help user build a dataset

TODO:

[x] make hint checking progress accessible for the user (needs the current count, the total count and an eta)
[x] pick a ubiquitous language (either hint, issues, checking)
[x] persist nlu hints on database
[x] add functions in nlu-client to call all hints routes
[x] add a janitor for hints and issues
[x] run hints computation in a dedicated process to prevent from freezing the main process (#159)
[x] add a route to cancel a hint computation + add that behavior (#159)
[x] ensure canceling a task when task queue is on database does not crash server
[x] add few tests for linting routes
[x] allow any in E_000
[x] make E_000 faster by only computing entities on tagged tokens
[x] allow specifying a speed in POST /lint route + if speed is 'fastest' compute hints on local instance, in main process
[x] rename e_004 for i_000 (it's an info as we simply want to indicate to the user that these are trimmed)
[x] fix e_000 to make sure consecutive space dont affect the algorithm

botpress / nlu