use immutable buffers in SubtleCrypto methods

[eric-seppanen]

This change makes all the methods in SubtleCrypto use &[u8] parameters instead of &mut [u8].

Here are the affected parameters:

• decrypt(data): the WebCrypto API states:

  Let data be the result of getting a copy of the bytes held by the data parameter passed to the decrypt method.

• digest(data): the WebCrypto API states:

  Let data be the result of getting a copy of the bytes held by the data parameter passed to the digest method.

• encrypt(data): the WebCrypto API states:

  Let data be the result of getting a copy of the bytes held by the data parameter passed to the encrypt method.

• sign(data): the WebCrypto API states:

  Let data be the result of getting a copy of the bytes held by the data parameter passed to the sign method.

• unwrapKey(wrapped_key): the WebCrypto API states:

  Let wrappedKey be the result of getting a copy of the bytes held by the wrappedKey parameter passed to the unwrapKey method.

• verify(signature, data): the WebCrypto API states:

  Let signature be the result of getting a copy of the bytes held by the signature parameter passed to the verify method.
  Let data be the result of getting a copy of the bytes held by the data parameter passed to the verify method.

I take the API documentation language "getting a copy of the bytes" to mean that the original buffer is left unmodified.

I can't find any evidence that these buffers would be modified. I'm not sure how strong an argument that is.

Fixes #3795.

[daxpedda]

LGTM!
Can you add an entry in the changelog?

[eric-seppanen]

Added a changelog entry.

Rebased due to merge conflicts in the changelog; I made no other changes from the previous version.

[daxpedda]

Thank you!

[eric-seppanen]

I realize after the fact that this change made the 0.3.68 release not semver-compatible, because the signatures of some SubtleCrypto methods changed. Example: 0.3.67, 0.3.68.

Apologies for the breakage! Perhaps cargo-semver-checks could have caught this?

[daxpedda]

AFAIK changing a parameter from &T to &mut T should not be a breaking change.

Is that what you meant or did I miss something else?

[eric-seppanen]

@daxpedda Oops, you're right. (Though it's &mut T -> &T). I confused myself by copying some code from a workspace using 0.3.68 into another project using 0.3.67 and it didn't compile. But that's not semver breakage, it's just that downgrade doesn't work. Sorry for the noise.