TaskState: remove cached prereq satisfaction in preparation for cylc remove

[MetRonnie]

These manually cached satisfaction states have the possibility to drift out of sync with individual Prerequisite satisfaction states, especially with cylc remove work in progress #5643 (cylc remove will have to unset any satisfied prerequisites that the removed task provided).

It looks like they are barely used in the first place (at least not anymore):

• _is_satisfied only has any savings at this call to TaskState.prerequisites_are_not_all_satisfied():
  

  cylc-flow/cylc/flow/task_pool.py

  Lines 1755 to 1760 in 80fa008

  	# Satisfy any absolute triggers.
  	if (
  	itask.tdef.has_abs_triggers
  	and itask.state.prerequisites_are_not_all_satisfied()
  	):
  	itask.satisfy_me([

  which is the case for absolute triggers e.g. foo[2001] => bar which are not so common?

• _suicide_is_satisfied only has any savings at the call above and this call:
  

  cylc-flow/cylc/flow/task_pool.py

  Lines 1425 to 1430 in 80fa008

  	# Event-driven suicide.
  	if (
  	t.state.suicide_prerequisites and
  	t.state.suicide_prerequisites_all_satisfied()
  	):
  	suicide.append(t)

  suicide triggers are also not so common.

I am not sure this is strictly necessary for cylc remove (but not happy about the current situation), hence putting it up for review standalone first.

These cached values were first introduced in #1775. There was an issue #1857 mentioning it - looks like it was the cached states drifting out of sync that caused the problem in that case, which is why I am unhappy about it generally.

Check List

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• No dependency changes
• No tests needed
• No changelog needed as not visible to users
• No docs needed
• On master branch as code refactor

[oliver-sanders]

Makes sense, we have always assumed that prereqs cannot revert to unsatisfied after being satisfied, however, cylc remove will need to do this.

I expect this cache was introduced for efficiency reasons, may be worth checking that it doesn't perform poorly for tasks with large numbers of prereqs (though I'm not expecting much change by the looks of the code, as you say they are barely used).

[oliver-sanders]

Tested with the "Diamond" workflow. CPU usage actually went down slightly as a result of taking out the caching. It definitely wasn't doing anything useful.