More lalrpop

[wackywendell]

This builds on #64:

• Migrate extension declaration parsing to LALRPOP — replace hand-written parsers with grammar rules and structured AST types, consistent with how relations are already parsed
• Extract extension lowering into its own module — separate extension-specific lowering from relation lowering, with shared validation helpers
• Idiomatic trait usage — add FromStr and From impls for extension types to clean up conversion boundaries
• Documentation — add module-level docs across the codebase, document the extension parsing pipeline, and add a CI step for docs compilation
• Visibility cleanup — tighten Lower trait to pub(crate) and simplify imports

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8c33589909

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Reject oversized extension anchors without panicking

The new ExtensionUrnDecl/ExtensionDecl grammar routes extension anchors through Anchor/UrnAnchor, which currently convert with parse::<u32>().expect(...); inputs like @4294967296: ... or #1 @4294967296: fn will now panic the parser process instead of returning a structured parse error. Before this migration these declaration lines used fallible parsing (map_err) and surfaced normal MessageParseErrors, so this introduces a crash-on-malformed-input regression.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Broaden URN tokenization for extension declarations

The new URN_VALUE lexer rule only accepts URL-like scheme://..., absolute /..., identifiers, or quoted names, so URNs such as urn:substrait:functions_arithmetic (previously accepted because everything after : was treated as URN text) now fail to parse. This is a behavioral regression in extension parsing and can break parse/textify round-trips for plans containing non-URL URN formats.

Useful? React with 👍 / 👎.