[Feature Request] Overload CEL matches function to support bytes values

[rodaine]

Feature description:

CEL supports <string>.matches(<string>) to apply an RE2 regular expression to strings. We expose this functionality in the custom/predefined constraints, as well as the underlying implementations of strings.pattern and bytes.pattern standard constraints.

Since the matches function only targets a string receiver, bytes.pattern first converts the value to a string before applying the regex. This only works if the bytes are a valid UTF8 string, failing with a runtime error within the CEL expression if it's not. This prevents the use of arbitrary byte sequences with the with regex.

To combat this, matches should be overloaded to support <bytes>.matches(<string>) which applies the regular expression without casting the receiver.

Problem it solves or use case:

Proper application of regular expressions against non-UTF8 bytes fields.

Proposed implementation or solution:

Add an overload to the custom CEL functions protovalidate requires of <bytes>.matches(<string>). This change can/should be upstreamed to the CEL project as well.