gh-141968: Use take_bytes in zipfile._ZipDecrypter

[cmaloney]

Removes a copy going from bytearray to bytes.

• Issue: Remove bytearray to bytes copies in stdlib using take_bytes #141968

[cmaloney]

@serhiy-storchaka would you be up for taking a look at this?

[serhiy-storchaka]

Does it cause a measurable difference?

[serhiy-storchaka]

I do not think this has a measurable effect, but at least it does not make it worse.

[cmaloney]

Keep hunting for a zipfile to test with that's 1KB or larger and encrypted this way but haven't found one / not a commonly used thing in the infra I'm familiar with.

In theory the amount of time saved is proportional to the memcpy time on most platforms. The core interpreter loop in my measuring is fast for most these I've done and the perf delta ends up being 5-10%...

This particular one there is some argument we could pattern recognize at the JIT level: If a bytearray is uniquely referenced in a call to bytes in a return (a point where no further references could be created) then can take the buffer / convert to a .take_bytes(). That seems a long ways off and this still wouldn't pessimize / is expressing intent "the result bytearray is no longer owning this storage)