"Nitpick" linters

[SquidDev]

It'd be really nice to have a series of linters which nitpick about minor formatting. For instance, ensuring spaces are in the correct places, enforcing a variable name convention, etc... Ideally many of these could have automatic fixers too, allowing for a form of automatic error correction.

• [x] Require spaces around binary operators, and assignments.
• [ ] Require spaces after a comma, and forbid any before.
• [x] Enforce a consistent table separator (defaults to ,, but allow ;).
• [x] Require trailing ,/; on multi-line tables.
• [ ] Complain about trailing whitespace. This may be a little harder, as it'd require providing a zipper-list of all tokens (or at least a "next token").
• [ ] Misleading/inconsistent indentation.
• [ ] Enforce variable naming conventions for locals and table keys/methods. It might be worth considering what we do with something like CC:T which uses camelCase for functions (even if local ones), and snake_case for variables.
• [x] Consistent spacing within parenthesis and {} (i.e. forbid {foo } but allow {foo} and { foo }).
• [ ] Consistent table key names (forbid { ["foo"] = 1, bar = 2 }).
• [x] Convert string indexing into dot ones (x["foo"] → x.foo). Might be nice to have a reverse refactoring too when we work that out.
• [x] String quotes (" vs ')
  • [ ] Allow "prefer " and prefer '". Prefer a kind of quote, but allow the other when you have strings like 'foo "bar"'.
• [ ] Remove redundant =s in long strings.
  • [ ] And comments
• [ ] Convert long string to quoted ones when nothing need be escaped. Might be nice to do the other way as a standard refactoring too.

One thing bearing in mind is how to handle alignment within tables and assignments. For instance, ideally we'd require a single spaces around tokens, but allow for cases such as follows, where people are trying to align expressions up:

vx = vx + 1
x  = x  + 1

[SquidDev]

We check trailing whitespace on commas on 1a7300d7d05d0ef6f732f99006e04f833211abb0, but doing it generally is going to be harder. The token linter pass really needs a way to get the previous and next node, or at least their whitespace.