refactor: remove ProxyRuntime and add proxy feature to PocketIcRuntime

[lpahlavi]

This PR removes the ProxyRuntime added in #84 and instead adds a proxy feature to the PocketIcRuntime to optionally route update calls through a proxy canister.

[gregorydemay]

Thanks for this PR @lpahlavi ! Only some minor comments

[gregorydemay]

a couple of nits (could be for another PR)

1. We could re-use target to store wasms, e.g. target/test-artifacts/<name>.wasm
2. This logic is not the same as the one running locally. Apparently (according to chatgpt) a standard solution there would be to create an xtask, e.g. cargo xtask fetch-proxy-wasm which can be run locally and on CI. See here. Like this also the version needs to be defined only once.

[lpahlavi]

We could re-use target to store wasms, e.g. target/test-artifacts/.wasm

Good idea, done!

This logic is not the same as the one running locally. Apparently (according to chatgpt) a standard solution there would be to create an xtask, e.g. cargo xtask fetch-proxy-wasm which can be run locally and on CI. See here. Like this also the version needs to be defined only once.

Cool! I didn't know about xtask, I'll have a look and do this in a separate PR 💪

[gregorydemay]

nit: reqwest is quite heavy in terms of dependency, why not just call curl? E.g.

let path = "target/test-artifacts/module.wasm";

    if !std::path::Path::new(path).exists() {
        std::process::Command::new("curl")
            .args([
                "-L",
                "-o",
                path,
                "https://example.com/module.wasm",
            ])
            .status()
            .unwrap();
    }

[lpahlavi]

Good point! Done.

[gregorydemay]

This looks scary, can be simplified with

Suggested change

	C: CyclesChargingPolicy + Clone,
	<C as CyclesChargingPolicy>::Error: std::error::Error + Send + Sync + 'static,
	C: CyclesChargingPolicy<Error: Into<BoxError>> + Clone,

(because BoxError is just an alias for Box<dyn Error + Send + Sync>)

[lpahlavi]

Ah I forgot I could use Into<BoxError>... Done!

[lpahlavi]

I also didn't know I could use this shorthand syntax to combine the two bounds in a single line!

[gregorydemay]

nit: It's a bit weird that this method is declared to return Vec<u8> when it's actually the never type !. Since it's only used in one place, it would be simpler to inline it, no?

[lpahlavi]

Good point. Done.