added treesitter checking for codeblocks

[KirkaigDev]

First file I used this plugin on was fairly big and the regex checker caused the header search to fail somehow. This fixes that.

[jakewvincent]

Hi @KirkaigDev, thanks for the contribution! Could you share a file, or a smaller example in a comment here, that reproduces the heading navigation failure, plus any steps needed to trigger the failure? I’d like to understand first where the existing code block detection is going wrong.

I’ve intentionally kept mkdnflow’s navigation independent of treesitter since we can’t assume users have treesitter or the Markdown parser installed. I see that you included a fallback, but the unconditional treesitter call can fail before even reaching that fallback if the required parser isn’t available.

Generally I’d prefer to identify a fix to the existing detection before introducing treesitter into navigation. I've considered adding treesitter integration and am not opposed to it, but it would need to be opt-in, with parser availability checks, health diagnostics, and working alternative non-treesitter paths. But that’s a bigger change than I’d want to take on as part of this bug fix.

For now I'm thinking we can start with a reproduction and see if the failure can be addressed without adding that integration. When you can, please share the info I'd need for a minimal repro. Thanks!

[KirkaigDev]

My bad, I forgot to wrap it in a pcall. It should work now. I don't use a lot of markdown so I can't find a smaller file, but the error occurred here https://github.com/neovim/nvim-lspconfig/blob/master/doc/configs.md

[KirkaigDev]

I figured out what caused the issue, it's because there was a leading space in one of the code blocks. It seems like an easy fix to just change the regex matching to '^ *```' but somehow that doesn't work. No idea why, just removing the space fixed the entire file, but for whatever reason I can't get the regex to fix it and yet treesitter does fix it.