Phase 7: the CLI knows who you are and where you wrote it

Phase 7: the CLI knows who you are and where you wrote it

Three things a first-time user hit in a row, all measured against a live
instance before they were fixed.

1. Identity was guessed. Every issue and request was stamped with a name from
   `git config user.name`, sanitised: "Keenan Nemetz" became `keenan-nemetz`, an
   account that does not exist, and nothing complained — the issue page showed it,
   and the assignee rule accepted it because that rule only checks the shape.
   `vole-whoami` is a new SSH verb, and `vole whoami` is the CLI command that
   reads it: the instance is the only place a key maps to an account, and it is
   one call on the connection already open. `vole issue` and `vole request` stamp
   the account's current name and id, and `--author` is gone from both, because a
   name you can type is a name you can get wrong. Read commands still need no
   identity — `issue list` and `issue show` keep working with the instance
   unreachable — and the git-config guess survives only as a warned fallback for
   `--no-push`.

2. `--remote ssh://…` did not fetch. `fetch` checked for a configured remote name,
   so a URL was skipped: the first write in a fresh repository succeeded (nothing
   to fast-forward from) and every later one pushed from a tip that had never been
   updated, was rejected, and then retried as a "lost race" that never happened,
   reporting `already exists`. A URL is now fetched like a named remote, so
   `git init` plus a URL is the real no-clone filing path the roadmap described.

3. A command answered with an id and stopped. `new` and `comment` now print the
   page they wrote, and a comment link lands on that comment: the instance reports
   its own `public_url` in the same whoami reply (new optional config key), and the
   web puts an `id="c-<ulid>"` anchor on every comment. Empty states on the issues
   and requests pages print the command that fills them, because four of the five
   things a reader wants to do are writes the web cannot take, and "no comments
   yet" is a dead end.

Verified live, not only in tests: `vole whoami` in a clone whose `user.name` is
"Keenan Nemetz" printed `keenan` while the issue it then filed carried
`author: keenan` on the branch; two repositories that never clone both wrote
through the URL (the exact case that failed before); and a comment printed a
`#c-01M5N3M0…` permalink whose anchor is on the served page. Tests:
TestWhoAmIAnswersFromTheInstance (including `--author` now being an unknown
option), TestReadCommandsNeedNoIdentity (closes the SSH listener, then lists),
TestRequestAuthorComesFromTheKey, TestAURLRemoteIsFetchedBeforeWriting,
TestEmptyStatesPrintAWorkingCommand, and the comment-empty-state assertion added
to TestIssuePushedThenRendered. scripts/ci.sh green (gofmt, vet, build, markdown
links, go test -race ./...).