fix(FhenixHRE): handle hex-encoded ciphertexts correctly in hardhatMockDecrypt

[amathxbt]

Summary

hardhatMockDecrypt in FhenixHardhatRuntimeEnvironment.ts converts a ciphertext string to a bigint by splitting it into characters and calling charCodeAt(0) on each, treating the result as a byte value. This is incorrect for hex-encoded ciphertexts (the format returned by FHE sealoutput on real networks) and for any byte value > 127.

Bug

// Before
function hardhatMockDecrypt(value: string): bigint {
  const byteArray = new Uint8Array(
    value.split("").map((c) => c.charCodeAt(0))
    //                            ↑ returns a Unicode code point, not a byte
    //                            For a hex string "0x1a2b", this would encode
    //                            ['0','x','1','a','2','b'] = [48,120,49,97,50,98]
    //                            instead of [0x1a, 0x2b]
  );
  ...
}

Consequences:

1. Any hex-prefixed ciphertext (e.g. returned by a forked network or a mock sealoutput that emits hex) is decoded completely wrong, producing an arbitrary bigint that does not correspond to the original plaintext.
2. Bytes > 127 are silently handled as their Unicode code point (same as the byte value for ASCII, but the original code lacked the & 0xff mask, which is a portability hazard when the JS engine returns a multi-byte code point).

Fix

Detect hex-encoded inputs (0x-prefix or valid even-length hex string) and parse directly with BigInt("0x..."). Fall back to the byte-array path for plain mock strings, now with an explicit & 0xff mask for safety.