Fix spurious 'can't reuse name' errors when a multi-file package has a syntax error

[pmetras]

Fix the problem demonstrated into issue #4160: getting extra error messages in situations where one .pony files of an included (used) module has a syntax error and another of the module files that the main code is importing has multiple for loops using the same loop variable.

The issue was caused by the compiler's package loading logic. When a package containing multiple files was loaded, if one file had a syntax error, the package was left in a partially-processed state. Subsequent attempts to use or "catch up" that package to the current compiler pass would re-run passes (like the scope pass) on the existing AST. Since the scope pass had already partially populated the symbol tables, re-running it caused it to encounter its own previously defined symbols and report them as illegal name reuses.

The fix is to set AST_FLAG_PRESERVE on the package when parse_files_in_dir fails (same pattern already used when ast_passes_subtree fails). This stops future passes from visiting the broken package.
The magic->mapped_path branch also needed the same guard, because add_package_path uses it and parse_files_in_dir has the same partial-parse behavior in both branches.

Added BadPonyTest.SpuriousErrorMessageGeneration test with the new add_package_path utility method to be able to test on multi-files packages.

[ponylang-main]

Hi @pmetras,

The changelog - fixed label was added to this pull request; all PRs with a changelog label need to have release notes included as part of the PR. If you haven't added release notes already, please do.

Release notes are added by creating a uniquely named file in the .release-notes directory. We suggest you call the file 5147.md to match the number of this pull request.

The basic format of the release notes (using markdown) should be:

## Title

End user description of changes, why it's important,
problems it solves etc.

If a breaking change, make sure to include 1 or more
examples what code would look like prior to this change
and how to update it to work after this change.

Thanks.

[SeanTAllen]

The comment here ties the explanation to for-loop sugar, but the underlying problem is more general. Any pass running on a partially-processed package could produce spurious errors. The for-loop case is just how this manifested in #4160. I'd drop the for-loop detail and keep it general — "producing spurious errors in those files" is enough. The for-loop specifics are better suited to the commit message where they're tied to the issue report.

[SeanTAllen]

Could use a brief comment here. Even just a pointer to the comment on the other branch below so someone reading this code path doesn't have to go hunting for why the flag is being set.

[SeanTAllen]

paragraphs in release notes shouldn't have linebreaks. they should flow to fit the space they are in,

[SeanTAllen]

The release notes in this PR contain implementation details that aren't relevant to users. Here's a suggested replacement:

## Fix spurious 'can't reuse name' errors when a multi-file package has a syntax error

When a package contained multiple `.pony` files and one of them had a syntax error, the compiler would produce spurious `can't reuse name` errors for unrelated valid code in the other files. This was most visible when a valid file reused a loop variable name across consecutive `for` loops — a perfectly legal pattern that the compiler would incorrectly flag.

The spurious errors no longer appear. You'll still see the real syntax error, but the compiler won't pile on with bogus errors from other files in the same package.

[SeanTAllen]

please remove the hard wrapping in this file.

[SeanTAllen]

we don't do any line breaks in release notes except for paragraphs so they can flow to the viewport in the various places they are displayed.

[SeanTAllen]

im not a fan of adding fixtures to a testing style we are trying to move away from. @jemc what are your thoughts?