Implement (if-let) guards in pattern matchings

[maximebuyse]

Closes #814

This adds support for guards in pattern matchings. They are added to the AST (with a corresponding feature guard), and handled in the THIR import. A phase rewrites them as an equivalent nested pattern matching. Apart from supporting guards as specified in #807, this will also be useful for future phases that will generate guards (see #806).

Example of code with if-let guard (which is currently a nightly feature) and Rust code for the equivalent after the new phase:

pub fn if_let_guard(x: Option<Result<i32, i32>>) -> i32 {
    match x {
        None => 0,
        Some(v) if let Ok(y) = v => y,
        Some(Err(y)) => y,
        _ => 1,
    }
}

pub fn equivalent(x: Option<Result<i32, i32>>) -> i32 {
    match x {
        None => 0,
        _ => match match x {
            Some(v) => match v {
                Ok(y) => Some(y),
                _ => None,
            },
            _ => None,
        } {
            Some(y) => y,
            None => match x {
                Some(Err(y)) => y,
                _ => 1,
            },
        },
    }
}

Regular if guards are also supported (but for now desugared as if-let guards):

pub fn if_guard(x: Option<i32>) -> i32 {
    match x {
        Some(v) if v > 0 => v,
        _ => 0,
    }
}

pub fn equivalent(x: Option<i32>) -> i32 {
    match x {
        Some(v) if let true = v > 0 => v,
        _ => 0,
    }
}

[maximebuyse]

This also closes #818 and #807

[W95Psp]

Thank you, the output looks like what we want!
However, we need:

• more tests (e.g. a match where more than one arm as an if/if let, exercise if, exercise a non-trivial if let with a computation in the RHS... you don't need to have the equivalent desugar for everything)
• some description on the PR. Your PR should:
  • explain the problem you solve (referring to an issue is fine, iff the issue has a nice and intelligible description about the exact problem we are solving, with examples --but in your case the issue has no description, please add some as requested in the issue)
  • then give an overview of technically what is our approach
  • then give examples of what one can do if your PR is merged (possibly with links to playground)
  • note that those points above should basically be copy pastes from code documentations or tests cases

[maximebuyse]

Thank you, the output looks like what we want! However, we need:

* more tests (e.g. a match where more than one arm as an `if`/`if let`, exercise `if`, exercise a non-trivial `if let` with a computation in the RHS... you don't need to have the equivalent desugar for everything)

* some description on the PR. Your PR should:
  
  * explain the problem you solve (referring to an issue is fine, iff the issue has a nice and intelligible description about the exact problem we are solving, with examples --but in your case the issue has no description, please add some as requested in the issue)
  * then give an overview of technically what is our approach
  * then give examples of what one can do if your PR is merged (possibly with links to playground)
  * note that those points above should basically be copy pastes from code documentations or tests cases

Thank you. I addressed most of your feedback but I can add more tests / documentation if you think it is not clear enough.

[W95Psp]

Thanks you, that's great! :)
I don't know how you want to do this optimization for "one wildcard arm" matches: in this PR or in a subsequent one?

[maximebuyse]

Thanks you, that's great! :) I don't know how you want to do this optimization for "one wildcard arm" matches: in this PR or in a subsequent one?

I'll do it in this one

[maximebuyse]

Thanks you, that's great! :) I don't know how you want to do this optimization for "one wildcard arm" matches: in this PR or in a subsequent one?

@W95Psp That's done now

[W95Psp]

Thanks, I think that's great now! Let's merge that PR :)