Performance audit/improvements

[jmrog]

It would be good to buckle down a bit and make sure that we're doing things as efficiently as possible. Some example areas where there's likely room for easy improvement:

1. It likely makes sense to code-split each parser into its own bundle, rather than always bundling them all together. This will improve loading times. Covered by Split each parser into a separate bundle #31.
2. Our utils/regex helper always uses capturing groups, but this is almost never necessary. Non-capturing groups should be more performant.
3. The visitor in the SrsParser likely visits more nodes than necessary; there are probably opportunities to short-circuit.