fix: use cargo-metadata to support init command in workspace

[0xalpharush]

closes #333

Instead of manually parsing the toml file and failing in workspace members, the package name and edition are obtained via cargo-metadata (this also seems to obviate the need to pass that path which is a breaking change to a public API). The way the tests are setup was making cargo-metadata think that each test is a member of cargo-fuzz so I added an empty workspace entry to the tests' Cargo.toml.

[fitzgen]

I've had issues with cargo-metadata blowing up compile times in the past. Can you provide the timings for cargo clean; cargo build both on main and on this feature branch?

[0xalpharush]

Using hyperfine on my M1, it adds ~1s

Feature branch:

Benchmark 1: cargo clean; cargo build
  Time (mean ± σ):      6.016 s ±  0.043 s    [User: 21.272 s, System: 2.145 s]
  Range (min … max):    5.968 s …  6.093 s    10 runs

Main branch:

Benchmark 1: cargo clean; cargo build
 Time (mean ± σ):      5.009 s ±  0.113 s    [User: 14.871 s, System: 1.736 s]
 Range (min … max):    4.859 s …  5.199 s    10 runs

[fitzgen]

One second doesn't seem too bad, given that this fixes bugs and cleans up some fiddly bits.

Some questions below.

[fitzgen]

Thanks!

[fitzgen]

Looks like CI is failing? Don't have time to dig in myself at the moment.

[0xalpharush]

The tests were relying on the panic error format pre rustc 1.73.0 https://blog.rust-lang.org/2023/10/05/Rust-1.73.0.html

[fitzgen]

Thanks!