docs: prose pass over README and published docs

[lesnik512]

Prose pass over the README and the published docs (everything mkdocs builds; docs/agents/ and docs/adr/ are excluded and untouched). Code blocks, inline code, commands, link targets and admonition syntax are unchanged; a script diffed every fenced block, inline-code span and link target against main. The only inline-code span removed is faststream-outbox inside a cut closer in comparison.md.

What changed

• Em dashes in prose: 317 to 0, replaced with commas, colons, parentheses, periods or rewrites. The 11 left are inside code blocks (code comments).
• Bold run-in labels and decorative bold: 117 lines to 0. The troubleshooting page accounted for 61. Its "Symptom / Likely cause / Diagnose / Fix / Reference" entries are now plain paragraphs, and every symptom, cause, diagnostic step and fix is kept. Other pages had **TL;DR.** labels, bold-led checklist and list items, and emphasis bold.
• Contrasts and closers: a few restating closers and staged openers were cut (for example "That atomicity is the whole point.", "That's the whole thing.", "The reason is simple:"). Contrasts where both halves carry information were kept.
• docs/index.md: decorative → arrows and the italic *concept:* / *reference:* / *step-by-step:* labels in the page index are gone.
• dlq.md: one sentence that read as version history ("Default behavior is unchanged when...") now describes current behavior.
• Headings: 8 headings changed a dash to a colon (and moved to sentence case). The default toc slugify drops both characters, so every anchor slug is unchanged. I checked this in the built HTML.

Social card

docs/assets/social-card.png is replaced with the card regenerated from the current repo description in modern-python/.github#125.

Checks

• just lint-ci: pass
• pytest against a throwaway postgres:17 container: 624 passed
• mkdocs build --strict: pass

[github-actions]

Benchmark gate

✅ gate passed

scenario	msg/s	delete/msg	WALrec/msg	WALB/msg	fpi	upd	del	dead_tup
consumer/w1/b10	929	1.000	6.79	921	242	5000	5000	10000
consumer/w1/b100	929	1.000	6.83	920	243	5000	5000	10000
consumer/w2/b10	1125	1.000	6.79	957	242	5000	5000	10000
consumer/w2/b100	1302	1.000	6.72	954	243	5000	5000	10000
consumer/w4/b10	995	1.000	6.82	995	261	5000	5000	10000
consumer/w4/b100	1502	1.000	6.87	1037	244	5000	5000	10000
consumer/w1/b100/tfbs100	4828	0.010	6.07	1116	243	5000	5000	10000
producer/w1/b100	2157	0.000	3.04	584	0	0	0	0

Gated (fails the build): delete_calls + tuple counters (upd/del/ins) + the producer's insert_calls, exact; select_calls within +2; wal_records within a 10% band. msg/s, WAL bytes and total calls are informational (timing/FPI noise).