Fixed mistakes in the canonicalization regular expressions and altered the
functions to use a source and destination buffers. This reduces the amount
of memory allocations that take place.
Added more canonical query tests to the internal test suite.
The query does not need to be parsed for it to be canonicalized and the parsing
uses the PCRE2 library. The regular expressions were changed so that only one
call to the function which replaces literal unquoted values is made.
