wip: stacktrace: enable method signature printing for inlined frames

[aviatesk]

This is an attempt to revive #41099 using the newly added invert linetable format (#52415).
It's a work in progress, but I've opened this PR to spark a discussion on whether I'm moving in the right direction with this approach. Currently, I've tweaked jl_lookup_code_address to give back a CodeInstance containing debuginfo rather than the MethodInstance for the outermost frame. This info is then used to find the MethodInstance for inlined inner frames from the Julia side.

julia> function f1(a)
           x = a
           return begin
               @inline f2(x)
           end
       end;

julia> function f2(a)
           x = a
           return @inline sin(x)
       end;

julia> f1(Inf)
ERROR: DomainError with Inf:
sin(x) is only defined for finite x.
Stacktrace:

diff --git a/old b/new
index f913ab20c3..215037aa56 100644
--- a/old
+++ b/new
@@ -1,10 +1,10 @@
  [1] sin_domain_error(x::Float64)
    @ Base.Math ./special/trig.jl:28
- [2] sin
-   @ ./special/trig.jl:39 [inlined]
- [3] f2
-   @ ./REPL[2]:3 [inlined]
+ [2] sin(x::Float64)
+   @ Base.Math ./special/trig.jl:39 [inlined]
+ [3] f2(a::Float64)
+   @ Main ./REPL[2]:3 [inlined]
  [4] f1(a::Float64)
    @ Main ./REPL[1]:4
  [5] top-level scope
    @ REPL[3]:1

[aviatesk]

One clear issue of the current implementation is we can't pinpoint the program counter for the outermost frame, making buildLineInfo unusable and causing lookups to fail except in the simplest cases. Is there a way, within lookup(pointer::Ptr{Cvoid}), to get a pc to feed into Base.IRShow.buildLineInfo(codeinst.debuginfo, nothing, pc)?

[vtjnash]

We could hijack the dwarf column to be the pc, or just entirely replace the line values with an index into our richer information stack