Handle error scenarios of HACL* functions

[chris-eibl]

This is a follow-up on #130960 (comment):

For all hash algorithms (NEW with this PR):
malloc, malloc_with_params_and_key, malloc_with_key: may return NULL (out of memory)
copy: may return NULL (out of memory)

Full list from #130960 (comment)

For Hacl_Streaming_HMAC:

• malloc may return OutOfMemory, InvalidAlgorithm (e.g. requesting Blake2b_256 on an ARM machine), or Success
• reset may return InvalidLength (if trying to reset the state with a key of different length, this is not supported), or Success
• update: MaximumLengthExceeded or Success
• digest: OutOfMemory or Success
• copy: may return NULL (indicates out of memory)

For all hash algorithms (NEW with this PR):

• malloc, malloc_with_params_and_key, malloc_with_key: may return NULL (out of memory)
• copy: may return NULL (out of memory)

For SHA3/Keccak only:

• digest may return InvalidAlgorithm (if the algorithm is shake)
• squeeze may return InvalidAlgorithm (if the algorithm is not shake)

I think all of these can be handled as a followup, I just thought it would be good to have it in writing here so that you can decide which of these are worth checking for. The reason I brought up other hash algorithms is that, since you requested (or maybe @picnixz ?) proper out of memory handling in HACL*, we now may return NULL for other algorithms (like hash algorithms), meaning that this PR will introduce new possibly-NULL return values as a side-effect of updating the vendored copy of HACL*.

For the record, Python ignores MaximumLengthExceeded on the basis that this cannot happen in practice.

Linked PRs

• gh-131316: handle NULL values returned by HACL* functions  #131324

[chris-eibl]

I don't think I've got enough bandwith atm. Please take over ❤