poc: remove Rc Cell from parser2 parse scope structs

[sbillig]

I've been wasting my time profiling the new parser. I noticed that it does a decent amount of allocating, and figured there may be easy speed wins. I'm fully aware that no one cares about this and I have better things to do with my life.

Here's an ugly poc of a minor improvement from a simple change; putting it here so that it doesn't fade away in my pile of git stashes.

Parsing a slightly modified uniswap contract:
739.23 µs -> 647.37 µs

(This pr is based on #860)

[Y-Nak]

Nice! This is Interesting.
I tried to run rust-analyzer's benchmark for its parser.
It takes around 20-40ms for 8000 lines file(you may need to modify the test code slightly, it might have a bug).
The command was

# in rust-analyzer/crates/syntax
RUN_SLOW_BENCHES=1 cargo test --release -- --nocapture --test-threads=1

Also, I tried running the benchmark on 9,000 lines file made by repeating uniswap-paraser2.fe about 20 times. It seemed to take about 20ms to execute.

So, it depends on the later passes, but I guess it should be fine for now.

[sbillig]

I agree, but if we remove the Rc Cell, it could take about 17ms! ;)