RFC: hashlib: allow getting digest without finalizing the hash object

[vshymanskyy]

During the implementation of OTA update, we came across a practical need to compute several (cumulative) digests of parts of a stream. The problem that we face is that the native implementation of the hashlib finalizes the hash object during digest() call.

We found 2 plausible ways to achieve what we need:

1. implement copy() method. This would allow us to use hash.copy().digest(), where only the copy of the hash is finalized, leaving the original hash intact.
2. make digest() operate on the copy of the state. With this approach, the original hash object does not need to be finalized.

Or, do both and make native hashlib implementations compatible with CPython. We have developed the PoC of (2) for SHA256 and it works like a charm. I would appreciate feedback on these approaches to determine if the uPy maintainers and community are open to accepting pull requests for these changes.

Note:

• These changes enable new features without breaking existing micropython code.
• The micropython-lib/hashlib does support copy() and also digest() operates on a copy of the state. However, pure python implementation is resource-hungry.
• Related to uhashlib not compatible with micropython-hmac #4500

@dpgeorge @pfalcon @jimmo