Rewrite lexer and parser and overhaul general syntax

[kevlarr]

This turned out to be a much larger overhaul than I initially set out to do. Originally it was about cleaning the existing lexer up so that it only concerns itself with valid tokens and not enforcing which tokens are expected next, but after coming back from a very long hiatus it rapidly turned into a "change all the syntax things" situation and, well, it's pretty comprehensive...

Notable changes include, in no particular order:

• Syntax is no longer whitespace-dependent, using () to open & close scopes and , to join column expressions, which allows for nicer things like record (col1 val, col2 val) and the removal of the very janky "indentation detection"
• Tables can be declared in the root scope and rely on search path for discovery, which resolves Support schema-less data files #32
• References are switched to always start with @ instead of having it in the middle, which resolves Qualified references look like email addresses #35 and eases tokenization
• Adds support for underscores in numeric literals, which resolves Underscores in number literals #48
• Multiline text strings and quoted identifiers are now officially supported and tested, which partially resolves Multiline text strings #41
• Groups named and anonymous records together, which resolves Preserve ordering of named and anonymous records #37
• Supports escaping single quotes in text strings, which resolves Support escaping single quote in normal string #31

Incidentally it also addresses the following bugs:

• Resolves Indented comment sets indentation unit #19

And last, and probably least, it resolves #50

TODO

• Rewrite lexer
• Rewrite parser
• Rewrite analyzer
• Rewrite loader
• Enrich lexer & parser error messages with positions
• Fix tests
• Review TODO and FIXME comments