Don't lower for codegen in functions that are never code generated

[TurkeyMan]

Scope.needsCodegen() is now false inside a function with skipCodegen, ie; a @__ctfe function, so its new is not lowered to the _d_newclassT hook, which a minimal runtime may not have.

[TurkeyMan]

This doesn't feel worthy of a changelog entry... it's really trivial.

[thewilsonator]

also check that without -c this fails compilation?

[TurkeyMan]

Not clear exactly what you mean; is this what you want to see? Ie, proof that it does not attempt to link _d_newclassT when ctfe new is used?

[github-actions]

DMD perf check

Metric	Base	PR	Δ
compile Phobos codegen (instr)	1,423.6 M	1,427.7 M	+0.292%

All measurements

Metric	Base	PR	Δ
compile hello.d (instr)	211.6 M	211.6 M	-0.004%
compile hello.d -O -release (instr)	229.2 M	229.1 M	-0.006%
compile Phobos (instr)	4,896.1 M	4,900.5 M	+0.089%
compile Phobos codegen (instr)	1,423.6 M	1,427.7 M	+0.292%
compile vibe.d (instr)	14,468.5 M	14,478.2 M	+0.067%
dmd binary size (stripped)	8.12 MB	8.12 MB	+0.05%
hello binary size (stripped)	0.72 MB	0.72 MB	0.00%
peak RSS (compile hello.d)	44.25 MB	44.22 MB	-0.07%
peak RSS (compile Phobos)	613.5 MB	613.6 MB	+0.02%
peak RSS (compile vibe.d)	1902 MB	1901 MB	-0.03%
page faults (compile hello.d)	8,933	8,931	-0.02%
page faults (compile Phobos)	153,372	153,373	0.00%
page faults (compile vibe.d)	477,606	476,928	-0.14%
compile dmd itself (wall)	12.5 s	12.5 s	-0.01%
compile hello.d (wall)	66.1 ms	63.6 ms	-3.81%
compile Phobos (wall)	1,517 ms	1,474 ms	-2.88%

_{c1f8c67 vs merge-base 10adc6c · about these metrics}