Better Handling of Candidates Who Become Invulnerable (#2801)

* remove candidate when to invulnerable

* fix

* candidates to collators

* make parameters consistent and more reasonable

* add call to kick invulnerable candidates

* factor removal into weight

* fix: use accrue instead of add

* make set_invulnerables non-atomic

* benchmark add_invulnerable to account for candidate removal

* don't remove from candidates with set_invulnerables

* fix bounds on benchmarking

* protect against zero min invulnerables underflow

* extra event and tests

* make candidates/invulnerables self-cleaning on session change

* add integrity test

* unused imports

* make rococo-contracts have 1 collator