Use acorn for parsing JavaScript code

[TerrorJack]

We used to use language-javascript for lexing or parsing JavaScript. This works for most cases, but the parser has rough edges (e.g. see #84 #86), and if we only do lexing, we don't have sufficient info about the AST (is it just an expression? is there top-level await?).

This PR switches to acorn for compile-time analysis of the inline JavaScript code in the quasiquoter. inline-js embeds a 108K bundled JS file, and use that for analysis. The source code for producing the bundle is also included.

The upsides of moving to acorn are:

• Much better support of latest ECMAScript syntax features. acorn is a much more battle-tested JavaScript parser.
• Much nicer API, one QuasiQuoter to rule them all. We used to have distinct expr/block quoters, then js/jsAsync. Now, we can finally only keep a single js quoter which supports all use cases.

The downside are:

• Calls node at compile-time. Can be bad for a variety of reasons: bad for cross compilation; may be risky; may be slow.

Anyway, let's make the switch since this is something I've wanted to do for a long time.

Also sneaks in some other changes:

• Use cabal for testing on Windows
• Misc CI-related bikeshedding

Close #84 #86