Feature request: HW accelerated crypto

[bugadani]

We should be able to parametrize embedded-tls to allow replacing software crypto with hardware implementation. This would be a giant perf win on certain MCUs, however, this seems quite a big overhaul.

We could probably base the API around https://github.com/RustCrypto/traits

[newAM]

This would be a good feature to have. I forget the context but I recall difficulties when I tried to use the STM hardware PKA for the RustCrypto traits.

For reference, I previously measured this for p256-cm4 on a STM32WL.

Implementation	Signing Cycles (appx)	Verify Cycles (appx)	Flash Size (appx)
Hardware PKA	5,211,859	10,516,860	1,582 B
RustCrypto	7,856,436	14,303,277	49 kiB
p256-cm4	442,754	1,225,746	10 kiB