feat(channels): adopt Telegram Bot API 10.1 rich-text formatting and fix guest-mode escaping

[bug-ops]

Summary

• Guest-mode responses (TelegramChannel::flush_chunks) now format through the same markdown_to_telegram renderer used by regular messages, sending parse_mode = "MarkdownV2" instead of an unconverted "HTML" call that previously either broke Telegram's HTML parser or rendered literal Markdown to guest users.
• TelegramRenderer now prefixes every line of a multi-line blockquote with > (previously only the first line was quoted). Nested blockquotes flatten to a single >-per-level instead of accumulating an unescaped >>, which Telegram's MarkdownV2 grammar rejects outright (400 can't parse entities, dropping the whole message). A MAX_BLOCKQUOTE_NESTING_DEPTH = 512 cap (same pattern as MAX_CHUNK_DEPTH from fix(index,core): add depth-cap guards against unbounded recursion DoS #6595) bounds tracked-mark memory for adversarially deep input.
• Adds TelegramConfig.expandable_blockquote_min_lines (default 10): blockquotes at or above the threshold render as Bot API 10.1's expandable (collapsed-by-default) form; 0 disables it unconditionally. Wired into --init and --migrate-config.
• Wires TelegramConfig.guest_mode/.bot_to_bot into create_channel — these fields were previously accepted but never threaded into the TelegramChannel builder, so Guest Mode and Bot-to-Bot were unreachable in a running agent even with both enabled (found during review; scope expanded with user approval since otherwise this PR's own guest-mode fix could never be live-verified).
• Fixes a latent crash the wiring change exposed: the Guest Mode proxy's TcpListener was never set non-blocking before being handed to tokio, which panics on recent tokio versions the first time guest_mode = true is actually reached.
• Updates sibling spec specs/007-channels/007-1-telegram-guest-mode.md and corrects specs/007-channels/007-3-telegram-rich-text.md §10's nested-blockquote edge case (the originally-specified "cumulative depth" approach was invalid MarkdownV2; flattening is correct).
• Documents the new config field and blockquote behavior in book/src/guides/telegram.md.

Closes #6541

Test plan

• cargo nextest run — zeph-channels, zeph-config, zeph, zeph-core all green (4543/4543 in the final developer run; reviewer independently re-ran and confirmed 4396 passed/0 failed/14 skipped)
• cargo test --doc — passing
• cargo +nightly fmt --check — clean
• cargo clippy --profile ci --workspace --all-targets --features "desktop,ide,server,chat,pdf,scheduler,testing" -- -D warnings — clean (independently re-run by reviewer)
• gitleaks protect --staged — no leaks
• New unit tests: guest/regular MarkdownV2 formatting parity, multi-line blockquote (2-9 lines), boundary at exactly min_lines, min_lines = 0 disables expandable, nested-blockquote flattening (depth 20 and depth 600-over-cap), special chars inside a blockquote line, create_channel wiring of guest_mode/bot_to_bot from config
• Existing markdown.rs/telegram.rs suites pass unchanged (no weakened assertions — diffed to confirm)
• Migration step 104 (migrate_telegram_expandable_blockquote_config) has dedicated behavioral tests (append/noop-no-section/noop-already-present/idempotent)
• Live test against a real Telegram bot (guest-mode @mention response, 12-line expandable blockquote) — playbook scenarios written in .local/testing/playbooks/telegram.md but not yet run against a live bot; tracked as a follow-up live-verification step, not blocking merge given the extensive unit coverage and doc-confirmed MarkdownV2 grammar checks performed during review
• Code-fence-inside-blockquote interaction (upstream Telegram parser behavior is genuinely undocumented/inconsistent here) is deliberately left as a characterization test pinning current output rather than a guessed fix, flagged in the playbook for live verification