Add --color setting to CLI.

[suaviloquence]

Fixes #719

There is already code to recognize the CARGO_TERM_COLOR environment variable for setting color policy in GlobalConfig::new(). This PR makes it so that the --color=... flag takes precedence over the env var, but if the CLI flag is not set, then the env var still applies if it is set.

I'm not super happy with needing the if let block that comes with set_color_choice in order to not overwrite the env var - I'd definitely be willing to refactor the env var detection into a different function or make some other change.

I wasn't sure if tests were necessary/how to test this. If you want tests, let me know and I can add them.

[obi1kenobi]

Looks great so far!

Re: testing, could you add some tests of the various --color options and env var settings into a new file in the tests directory? You can use existing tests like the ones in rustdoc_edge_cases.rs as inspiration for how to pass arguments, manipulate env vars, and check stdout / stderr to ensure the output includes / avoids color codes, as expected.

Ideally, we'd have both a "all lints passed" and a "some lints failed" run of cargo-semver-checks, and the test would ensure each of them produces the appropriate output (non-)colorization based on the env vars and CLI flags, with the appropriate precedence between them.

[pksunkara]

Tangential comment. Please disregard since it's only for future note.

The recommended SOTA color-related logic is to use the rust-cli/anstyle crates.

[suaviloquence]

Added some tests in the tests/color_config.rs. Right now it just tests for the presence/absense of the ANSI color sequence \u{1b}[ in stderr based on the flag/env var combination. I didn't do anything with auto because I'm not sure how to emulate a tty with assert-cmd. I also introduced the predicates crate as a direct dev dependency (it was already required by assert_cmd; I just put it in Cargo.lock) in order to test if the stderr contained the sequence vs. matching the whole input which wouldn't be static (because of timing, for one).

[obi1kenobi]

Nicely done, thank you! I'm going to slip in a couple of tiny tweaks for a couple of nitpicks I had, and mark for merge as soon as CI passes.

[obi1kenobi]

Added some tests in the tests/color_config.rs. Right now it just tests for the presence/absense of the ANSI color sequence \u{1b}[ in stderr based on the flag/env var combination. I didn't do anything with auto because I'm not sure how to emulate a tty with assert-cmd. I also introduced the predicates crate as a direct dev dependency (it was already required by assert_cmd; I just put it in Cargo.lock) in order to test if the stderr contained the sequence vs. matching the whole input which wouldn't be static (because of timing, for one).

Also, this explanation was great! It was very useful context that helped me understand the pull request much more easily. Thanks for putting it together!