Upstream workflows

I just stumbled across https://github.com/Julian/tree-sitter-lean/blob/259a2daf7a699cc047221f1e043e4d045fcd6be2/.github/workflows/ci.yml#L40-L45

Have you tested again with the latest tree-sitter version (0.27.0)? Much of the `generate` codepath was rewritten to prevent exactly this sort of catastrophic memory use which a number of outlier grammars used to(!) suffer from.

If the situation is fixed now (which I would be interested to hear if not), you could consider using the official upstream ci and release workflows (including query validation, see, e.g., https://github.com/tree-sitter-grammars/tree-sitter-query/blob/master/.github/workflows/ci.yaml), which would be a requirement for including this parser in nvim-treesitter down the road.

Start with .github/workflows/ci.yml at lines 40-45 and compare it with the linked upstream workflow. Test the generate path using tree-sitter 0.27.0, then assess the official CI and release workflows, including query validation. Done means the latest version's memory behavior is understood and the relevant upstream workflow coverage is adopted if appropriate.

Written by the indexing model from the issue text.