fix: Better errors for missing fn keyword

[michaeljklein]

Description

Using force(just(Token::EOF)) to ensure the whole file is parsed appears to hide the module error.

Problem*

Resolves #3768

Summary*

Before:

❯ cargo run build
error: Parser: expected an end of input but found #[aztec(private)]
  ┌─ /Users/michaelklein/Coding/rust/noir/test_programs/compile_failure/no_fn_keyword/src/main.nr:1:1
  │
1 │ #[aztec(private)]
  │ -----------------
  │

Aborting due to 1 previous error

After:

error: Unexpected main, expected one of fn, internal, open, pub, struct, unconstrained, Attribute
  ┌─ /Users/michaelklein/Coding/rust/noir/test_programs/compile_failure/no_fn_keyword/src/main.nr:2:1
  │
2 │ main() {
  │ ----
  │

Aborting due to 1 previous error

Additional Context

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [Exceptional Case] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[jfecher]

Pull Request / Validate PR title is Conventional Commit (pull_request_target) Failing after 3s

This is failing because PRs need to be prefixed with either feat:, fix:, or chore: with an optional note for the category they're for, e.g. feat(traits): Implement generic traits or similar. The categories aren't pre-set so you can choose anything for the parenthesis. Most of the time it is omitted though.

This is to automatically sort PR titles into the change log for releases.

[jfecher]

Looks good, a good next step since you mentioned exploring the parser more generally after this would be testing to see if we can recover parsing after the failed fn keyword or after similarly failing to parse other top level statements. Ideally we could skip to the next one. This could be difficult though, so I'm not saying this must be the next step.

It looks like CI is also failing for this PR because it needs a cargo fmt

[jfecher]

LGTM 👍