fix: Optimize parser

[jfecher]

Related issue(s)

Resolves #791

Description

Optimizes the expression_with_precedence parser which was the main cause of performance lost in the frontend.

The command used for testing was nargo prove p on the source project

fn main(x: bool) {
    constrain x;
}

Timing details are averages of 10 runs and the results are as follows:

           master     this branch
debug      1.758s        0.350s
release    0.470s        0.163s

So it is roughly a 5x speedup in debug and a 2.9x speedup on release.

For a more detailed flamegraph, see the graphs uploaded in #791.

Summary of changes

The old expression_with_precedence parser recursively called itself for each precedence level, but also called each child level twice when constructing expr <op> expr, leading to an exponential blowup. Note though that this is from the time it takes to construct the parser before running it, so this was a large fixed cost rather than one that scaled with program size. This PR simply adds a .clone() to reuse the first parser and make this function linear rather than exponential in time.

Checklist

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt with default settings.
• I have linked this PR to the issue(s) that it resolves.
• I have reviewed the changes on GitHub, line by line.
• I have ensured all changes are covered in the description.
• This PR requires documentation updates when merged.

[kevaundray]

Why is this marked as "doc needed"?

[jfecher]

Not sure, it was automatically tagged with that.