Support ALL circom InfixOp

[voltrevo]

Resolves #54.

This branch relies on an update to sim-circuit, so you might need to cargo update to make it work.

Adds all infix ops

Notably, this includes AIntDiv which is possibly different from ADiv. For our purposes, division generally means modular division. To reduce confusion, we should perhaps take the position that it always means modular division, since explicit integer division is now available if that's the intention. Currently sim-circuit implements division as integer division, so the two operations are the same.

In circom a / b is (modular) division, and a \ b is integer division.

Discord discussion thread.

Fixes running & debugging individual tests in VS Code (and possibly other IDEs)

Prior to this PR, running a test involved command line argument processing because building a circuit requires an Input, and Input::new reads the command line. This breaks VS Code's functionality for running individual tests because it passes arguments that are unrecognized by that system, which is intended for the CLI, not tests.

I was similarly caught by this issue in the wasm-support branch (demo here), since there is no command line in the browser. The changes to input.rs here are based on the work from that branch, so this will reduce the changes needed later for wasm support.

Now Input::new is a pure function (well except maybe some platform-specific Path gotchas), and main.rs instead uses input_processing::generate_input, which replicates the old behavior (including command line processing).

Possible underconstrained bug

During testing, I was struck by the way we ignore signal outputs that are not assigned any values. I added this test:

pragma circom 2.0.0;

template underConstrained() {
    signal output x;
}

component main = underConstrained();

Currently this just outputs vec![]. I'm guessing it should be an error instead, but I left it as a TODO.

Add test for x == x fixed by sim-circuit update

My test for all infix ops contains several examples of using the same input on both sides of the op. x == x may be a little weird but it's perfectly valid and meaningful in other cases like x * x. sim-circuit was incorrectly generating CyclicDependencyDetected for this, so I added a more specific test that covers the issue.

[brech1]

LGTM