Mainly by replacing `assert_eq!(sql, ast.to_string())` with a call to
the recently introduced `verified()` helper or using `parses_to()` where
the expected serialization differs from the original SQL string.
There was one case (parse_implicit_join), where the inputs were different:
let sql = "SELECT * FROM t1,t2";
//vs
let sql = "SELECT * FROM t1, t2";
and since we don't test the whitespace handling in other tests, I just
used the canonical representation as input.
There's no Token::Period in such situation, so fractional part (from sec) was silently truncated.
Can't uncomment the test yet, because parse_timestamp() is effectively
unused: the code added to parse_value() in 5abd9e7dec
was wrong as it attempted to handle unquoted date/time literals. One
part of it was commented out earlier, the other can't work as far as I
can see, as it tries to parse a Number token - `([0-9]|\.)+` - as a
timestamp, so I removed it as well.
1) Simplified the bit in parse_datatype()
2) Made sure it was covered by the test (the "public.year" bit)
2a) ...the rest of changes in the test are to fix incorrect variable
names: c_name/c_lat/c_lng were copy-pasted from a previous test.
3) Removed the branch from parse_pg_cast, which duplicated what
parse_data_type already handled (added in the same commit even
2007995938 )
