chore(licm): Identify untested code; refactoring; minor fixes

[aakoshh]

Description

Problem*

Trying to find pieces of logic not covered by unit or integration tests.

Summary*

Add more unit test coverage to LICM.

Refactor and fix some stuff:

• Remove Constrain and RangeCheck from can_be_hoisted_from_loop_bounds: they don't look at the bounds, and the conditions they check are redundant with can_be_hoisted.
• Remove Call from can_be_hoisted_from_loop_bounds: same reason as above.
• Remove BlockContext::can_be_hoisted_with_control_dependence: I thought the extra jumping around to call can_be_hoisted again with true was confusing, and it was clearer to see the calls to can_be_hoisted with false and true next to each other, with the latter being restricted by block_context.can_hoist_control_dependent_instruction().
• Changed can_be_hoisted to return Yes | No | WithPredicate, rather than having to call it twice with false and true; this way if it returns No we don't have to retry in futility, and I think it reads clearer.
• Return false for loops without an exit block from the header in is_fully_executed. It used return true, but two examples I found were 1) an actual infinite loop in regression_7103 and 2) a while true loop with a break created by the fuzzer to avoid infinite loops. The latter is clearly not in the spirit of the method, as it has a break.
• Handle loops with headers that have no invariant parameter in get_const_upper_bound, e.g. while <bool-variable>.
• Consider the loop header as an ancestor of the blocks in the loop for calculating is_impure.
• Fix the simplification of ConstrainNotEqual

Untested code

The following changes don't cause any test failure:

• can_be_hoisted_from_loop_bounds:
  • ArrayGet can be disabled
  • ArrayGet can return true instead of false
  • Constrain(..) | ConstrainNotEqual(..) | RangeCheck { .. } can be disabled
  • Call { func, .. } can be disabled
• can_be_hoisted:
  • Call / Purity::Pure: return false instead of true
  • Call / Purity::PureWithPredicate: return false instead of hoist_with_predicate
  • Call / purity None: return true instead of false
  • MakeArray: return true or false instead of function.runtime().is_acir()
• can_evaluate_binary_op:
  • Add | Mul | Sub when unchecked: true: return false instead of true
  • Div | Mod: return false instead of true when !value.is_zero()
• simplify_from_loop_bounds:
  • SimplifyResult::Remove: return false instead of true
• simplify_induction_variable:
  • Constrain: remove loop_context.no_break from condition
  • ConstrainNotEqual:
    • remove loop_context.no_break from condition
    • remove block_context.can_simplify_control_dependent_instruction() from condition
    • disable altogether
• simplify_induction_variable_in_binary:
  • false if upper_bound <= constant.inc() if we remove .inc()
  • false if lower_bound > constant if we return false_value instead of true_value
  • true if lower_bound >= constant if we return true_value instead of false_value
  • false if upper_bound <= constant.inc() if we return true_value instead of false_value

This is not an exhaustive list, just what I found by tinkering with conditions.

I'll open follow up PRs to add tests, but wanted to get this PR reviewed because of the refactoring it contains.

Additional Context

I used the following commands to look for failing tests:

cargo nextest run --cargo-quiet -p noirc_evaluator loop_invariant
cargo nextest run --cargo-quiet -p nargo_cli --test execute forcebrillig_false_inliner_i64_min --filter-expr "not test(ram_blowup)"

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Test Suite Duration'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 60d09ad	Previous: 8d5ae13	Ratio
test_report_zkpassport_noir-ecdsa_	2 s	1 s	2

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Execution Time'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 2be3e26	Previous: 8d5ae13	Ratio
private-kernel-inner	0.019 s	0.014 s	1.36

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[TomAFrench]

LGTM