Improve codegen, accuracy of inlining cost for unknown intrinsics

[ianatol]

This PR gives a more accurate cost to non-constant intrinsics, which were previously calculated as cost 1000, and correspondingly improves their codegen a bit using a runtime lookup table. My first time really touching our C level codegen logic, so there may be some flaws there.

Credit for this idea and help along the way: @Keno

An example:

Before:

test_ukint(f) in Main at REPL[3]:1
│ ─ %-1  = invoke test_ukint(::Core.IntrinsicFunction)::Any
2 1 ─ 1000 %1 = (f)(Main.Int32, 5)::Any                                     │
  └──    0      return %1                                                   │
Select a call to descend into or ↩ to ascend. [q]uit. [b]ookmark.
Toggles: [o]ptimize, [w]arn, [h]ide type-stable statements, [d]ebuginfo, [r]emarks, [e]ffects, [i]nlining costs, [t]ype annotations, [s]yntax highlight for Source/LLVM/Native.
Show: [S]ource code, [A]ST, [T]yped code, [L]LVM IR, [N]ative code
Actions: [E]dit source code, [R]evise and redisplay
Advanced: dump [P]arams cache.
 • ↩

After:

test_ukint(f) in Main at REPL[1]:1
│ ─ %-1  = invoke test_ukint(::Core.IntrinsicFunction)::Any
2 1 ─ 20 %1 = (f)(Main.Int32, 5)::Any                                       │
  └──  0      return %1                                                     │
Select a call to descend into or ↩ to ascend. [q]uit. [b]ookmark.
Toggles: [o]ptimize, [w]arn, [h]ide type-stable statements, [d]ebuginfo, [r]emarks, [e]ffects, [i]nlining costs, [t]ype annotations, [s]yntax highlight for Source/LLVM/Native.
Show: [S]ource code, [A]ST, [T]yped code, [L]LVM IR, [N]ative code
Actions: [E]dit source code, [R]evise and redisplay
Advanced: dump [P]arams cache.
 • ↩

[vtjnash]

@nanosoldier runbenchmarks("inference", vs=":master")

Not really a giant improvement, but it does take the time for this call from 35ns to 17ns, and similar improvement if it is declared as just Builtin as well:

z::Core.IntrinsicFunction = Core.Intrinsics.neg_int
@bprofile z(1)

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here.

[vtjnash]

@nanosoldier runbenchmarks("inference", vs=":master")

[nanosoldier]

Your benchmark job has completed - no performance regressions were detected. A full report can be found here.