Strip generated script tags from MDX output

[jpers1]

Summary

Strip generated <script> elements from MDX output after rehype-expressive-code runs.

Why

Fenced code blocks are valid supported content in notes, but rehype-expressive-code injects script elements into the generated output. That output is later rendered through React in MdxContent, which triggers a client-side warning because scripts inside rendered React component output are not executed.

This shows up as a runtime warning when serving content with fenced code blocks, even though the author content itself does not contain raw <script> tags.

Changes

• add a small rehype cleanup step that removes script elements from the generated tree
• run that cleanup after rehype-expressive-code

Scope

This PR is intentionally narrow:

• it does not change authoring syntax
• it does not remove fenced code block support
• it does not alter unrelated MDX rendering behavior

The goal is to keep the existing code-block rendering while removing the part of the generated output that React/Next warns about.

Validation

• compiled a fenced code block through the MDX pipeline
• confirmed the compiled MDX no longer contains a script element in its output

Notes

This is separate from Markdown table support because the issue and fix are unrelated. Tables require parser support; this issue is about generated code-block output from the rehype stage.

[jpers1]

Reviewer note: the script-node cleanup is intentionally after rehype-expressive-code and limited to <script> elements only. The goal is not to change code fence support, only to remove generated script tags that React/Next warns about when the compiled MDX is rendered on the client.

This should not regress copy-button behavior here because the repo already wires copy interactions separately in the client code rather than depending on inline script execution from the MDX output.

[janezd]

Closed in favour of #248.