Scoverage: fix broken warnings related to tail recursion and infinite loop detection

[anatoliykmetyuk]

Fixes a bunch of warnings broken by scoverage related to tail recursion and infinite loops detection. Warnings were broken by:

• Coverage modifying the trees to inject instrumentation, breaking assumptions by the loop detection and tail recursion phases.
• Coverage modifying the spans of the trees when instrumenting them, which breaks the positions at which warnings are reported.

The solution by phases affected:

InstrumentCoverage - phase that instruments the trees to inject coverage probes:

• Do not instrument literal booleans in if conditions - they are getting in the way of rewriting if statements with literal conditions into their respective then or else branches.
• Preserve the spans of the trees when instrumenting them, so that warnings are reported at the original position of the code.

CheckLoopingImplicits - phase that detects infinite loops in initializers:

• Ignore coverage probes as ones not capable of leading to an infinite loop.
• Strip the coverage instrumentation from RHS of defs before analyzing them for loops.

TailRec - phase that rewrites tail recursive calls into loops:

• Modified isInfiniteRecCall - the method that detects infinite loops in tail recursive calls - to allow coverage probes as ones not capable of breaking an infinite loop.
• In transform, modify the default getter index in recursive call argument position to peel through the coverage instrumentation to recover the original default-getter parameter index.

How much have you relied on LLM-based tools in this contribution?

Moderately, for codebase analysis and tracing.

How was the solution tested?

sbt "testCompilation --enable-coverage-phase"

[anatoliykmetyuk]

CI is red due to main being red. Will rebase once main is green again.

[SolalPirelli]

I guess this is a consequence of InstrumentCoverage having to be early because coverage needs to instrument code the user actually wrote rather than code the compiler produced after optimizations?

[anatoliykmetyuk]

Yeah, quite a lot of issues are a consequence of the coverage phase being early and breaking the assumptions of the later phases.