Improve error message for identifiers with leading underscore

[tetektoza]

When users write identifiers starting with underscore, (e.g.: let _c: u32), the parser produces an uninformative error.

Root cause of that is that lexer splits _c into two separate tokens, and the parser had no grammar rule matching the pattern, so LARLPOP generates a generir error listing what tokens it expected at that position (which happened to be '(' for tuple destructuring).

So, this adds new lexer token - InvalidLeadingUnderscoreIdent that catches any leading underscore next to identifiers.

Before:

Error [EPAR0370005]: expected an identifier, '(' -- found '_'
    --> \\?\C:\Users\tetektoza\source\repos\leo\test_underscore_project\test_underscore\src\main.leo:3:13
     |
   3 |         let _c: u32 = a + b;
     |             ^

After:

Error [EPAR0370047]: Identifier cannot start with an underscore
    --> \\?\C:\Users\tetektoza\source\repos\leo\test_underscore_project\test_underscore\src\main.leo:3:13
     |
   3 |         let _c: u32 = a + b;
     |             ^^
     |
     = Identifiers must start with a letter.

Resolves: #28755

[tetektoza]

Argh, I didn't see it before (and I should've checked first), but now after pasting before and after to the PR message, it looks like the error message has been improved on mainnet. 😅 Maybe the current one (which is different compared to the one from #28755) is enough and this PR could be closed and issue too, since it seems like there was some improvement in this error message. Not sure, but pasting my solution anyways.

[mohammadfawaz]

In https://github.com/ProvableHQ/leo/blob/mainnet/compiler/parser-lossless/src/lib.rs, we already have a function that checks identifiers. Is it better to do this check there? I'm not sure how I feel about inserting invalid regrexes in the tokenizer 🤔

Also, when adding new errors, we have this quirk that new errors should always be appended at the end of the errors file, so that old error codes don't change.

[tetektoza]

Point 1 - hmm, you're talking about check_identifier right? I took a look on it, but _c becomes two tokens at this point, Underscore and Identifier, so the function runs only on Identifier token. I can extend it, but it seemed counter-intuitive at this point, it's also possible to modify grammar, but probably with a ton of changes.

Point 2 - ahhh, good point, thanks. I can adjust it.

Let me know about point 1 and my first comment in the PR, maybe we don't need it anymore. Not sure.

[mohammadfawaz]

Point 1 - hmm, you're talking about check_identifier right? I took a look on it, but _c becomes two tokens at this point, Underscore and Identifier, so the function runs only on Identifier token. I can extend it, but it seemed counter-intuitive at this point, it's also possible to modify grammar, but probably with a ton of changes.

Oh I now undersatnd. Yeah then doing it in in the tokenizer is probably the way to go. Thanks!

[mohammadfawaz]

Suggested change

	msg: "Identifier cannot start with an underscore",
	msg: "Identifiers cannot start with an underscore.",

[tetektoza]

Done.

[tetektoza]

Also moved the error at the end of the file as discussed, removed a lot of changes with newly updated error codes.

[mohammadfawaz]

LGTM!

[mohammadfawaz]

Closing in favour of #28909