Update configuration to add default --enable-fpe-safety

[roystgnr]

clang 15 and 16 will generate floating-point-exception emitting behavior at -O2 by default, which breaks even the MetaPhysicL submodule's make check tests. I see the attraction of that (presumably they're packing N float instructions into N/4 or N/8 SIMD instructions, and then if N%4 or N%8 but the instruction is division there'll be some NaNs ignored at the end?) but we have a lot of users relying on trapping FPEs, so let's make the default behavior safe for them.

This ought to fix our newly-clang-15-using Test clang recipe.

[roystgnr]

Well, damn. error: overriding currently unsupported use of floating point exceptions on this target [-Werror,-Wunsupported-floating-point-opt] looks like it might actually be related.

[moosebuild]

Job Coverage on 01a843e wanted to post the following:

Coverage

	48d244	#3697 01a843
	Total	Total	+/-	New
Rate	61.98%	61.98%	-0.00%	-
Hits	67677	67676	-1	0
Misses	41517	41518	+1	0

Diff coverage report

Full coverage report

This comment will be updated on new commits.

[roystgnr]

And it's only on ARM mac that it's failing, so I can't reproduce locally either...

[roystgnr]

Okay - with that hack I'm seeing some expected failures and some weird IM failures but the fix is working where it's supposed to; I'm going to head back upstream to do it properly now.

[roystgnr]

And now what should be the proper fix is in. Cross your fingers.

[roystgnr]

This appears to have finally fixed new Linux clang without having broken ARM Mac clang, and that "Test mac" failure is some new environment issue, so if the rest passes I'm merging.

[roystgnr]

And that "Failed but allowed" on Test MOOSE is the network issues hitting a python test again. This should be good to go.