Transformation hexDecode should not allow badly encoded inputs, or documentation should be updated

[fzipi]

Describe the bug

This comes from this discussion: corazawaf/coraza#1253.

Technically, there is one test that accepts a string that could not be generated by hexEncode.

👉 Is this the expected behavior?

I think accepting a broken input is just prone to more errors. Don't know how this affects a possible chain of transformations.

[marcstern]

hexDecode should indeed check if characters are valid.
Question is what to do in case of an invalid one ...