Document er-macro-transformer and channel-timeout-exception?

[baijum]

Two shipped-but-undocumented surfaces, one PR — both are pure prose touching no engine code, in disjoint files.

er-macro-transformer — user-facing docs (KEP-0006 step 6)

• New page docs/guide/er-macros.md (nav: Guide → Language, right after Language Reference): the (form rename compare) convention, rename's referential transparency (fresh identifiers, definition-site resolution, same-identifier-per-invocation, datum-wide renaming), compare's binding-aware free-identifier=? semantics — "same binding or both unbound," the exact rule a syntax-rules literal matches by, refusing a shadowed use-site spelling — verified against src/expander.zig (erCompareFn) rather than the issue text, plus the pinned ER≡syntax-rules hygiene guarantee, a cond-expand feature test, and a "when to reach for it" section with an expansion-time-validation example. States plainly that this is not R7RS-large's syntax-case (KEP-0007) and does not re-litigate the sibling sub-libraries — links the dev notes instead.
• docs/conformance.md: new row in the "Extensions beyond R7RS-small's scope" table (KEP-0006, Final, shipped v0.22.0 — verified against the tag), and the syntax-case gap row now mentions the ER surface that does exist.
• Cross-links: docs/guide/language.md Macros section, docs/guide/index.md, docs/guide/srfi-support.md's SRFI-211 row, and the Next-link chain.

channel-timeout-exception? — document the timeout condition

• docs/procedures/extensions.md: full reference entry ({ #channel-timeout-exception } + index metadata, so the generated procedures index picks it up) with REPL-tested guard examples for both send and receive; channel-send/channel-receive prose updated — expiry with a timeout but no timeout-val raises a distinct, testable channel-timeout condition (message "channel-send: timed out" / "channel-receive: timed out"), the direct analog of SRFI-18's join-timeout-exception?, not a generic error.
• docs/guide/concurrency.md: new "Channel timeouts" subsection in the fibers section.

Verification

• Every code sample executed against the v0.25.0 core binary (KAAPPI_HOME=$(mktemp -d)), including the full cumulative channel-blocks replay of extensions.md.
• mkdocs build --strict passes (exit 0) — new page is wired into nav, all anchors resolve.
• The new concurrency.md block shifts the sweep's hardcoded block indices in scripts/sweep/sweep_guide.py; renumbered and the block added to the checked program. Sweep sections re-run green locally: concurrency (2 passed), language, extensions.md (63 checks).

Closes kaappi/kaappi#2392
Closes kaappi/kaappi#2396