Avoid cloning during token creation

[alamb]

Rationale

Inspired by @davisp's #1591 I was looking at Token::make_word and I noticed it made 2 clones (owned strings). Each copy slows things down so let's avoid them

Changes

1. Change Token::make_word to reuse its argument and thus save a clone

API Changes

This changes Token::make_word and Token::make_keyword to take String instead of &str. This would be a breaking API change. If it is a problem I can make owned variants (like Token::make_word_owned or something) 🤔

Performance benchmarks:

++ critcmp main alamb_faster_keyword_lookup
group                                                    alamb_faster_keyword_lookup            main
-----                                                    ---------------------------            ----
sqlparser-rs parsing benchmark/sqlparser::select         1.00      6.4±0.03µs        ? ?/sec    1.00      6.4±0.03µs        ? ?/sec
sqlparser-rs parsing benchmark/sqlparser::with_select    1.02     31.5±0.10µs        ? ?/sec    1.00     31.0±0.13µs        ? ?/sec

🤔 it appears not to make much/any difference

[alamb]

I was looking at our benchmarks, and they are very limited. If we want to pursue optimizing sqlparser more I think it would probably be a good idea to add many queries (like TPCDS and TPCH) into the benchmarks