Add fast, layer-merged forward and inverse NTT code. #3850

Job	Run time
lint-markdown-link	27s
Quickcheck (aarch64)	53s
Examples (macos-latest)	11s
Quickcheck (x86_64)	45s
Linting (ubuntu-latest)	1m 25s
Quickcheck C90 (aarch64)	1m 7s
Examples (ubuntu-latest)	16s
Quickcheck C90 (x86_64)	44s
Quickcheck lib (macos-latest)	13s
Quickcheck windows-latest	33s
Quickcheck lib (ubuntu-latest)	17s
CBMC (ML-KEM-1024)	0s
Functional tests (${{ matrix.target.arch }}${{ (((matrix.target.mode != 'native') && ', cross') \|\| '') }})	0s
CBMC (ML-KEM-768)	0s
CBMC (ML-KEM-512)	0s
Compiler tests (${{ matrix.compiler.name }}, ${{ matrix.target.name }})	0s
Non-standard configurations	0s
Compatibility tests (${{ matrix.container.id }})	0s
Platform tests (${{ matrix.target.name }})	0s
Compatibility tests (${{ matrix.container.id }})	0s
	6m 51s

Provide feedback