Sometimes the input goes through a bunch of processing before it is
passed to the actual code being fuzzed. For example, you may have a
process that converts the input into a pair of UTF-8 strings. When this
is the case, it's trickier to understand what the failing input was
if libFuzzer is only outputting the byte stream.
This commit adds LLVMFuzzerCustomOutput, which can be used to print
the failing input string with custom formatting.