docs(contributors): refresh contributor stats to v0.51.44

[nesquena-hermes]

What this is

Refresh of CONTRIBUTORS.md and the README contributors section to reflect the contributor reality at v0.51.44. The previous refresh was at v0.50.245 (~50 tags ago).

The numbers

	Before (v0.50.245)	After (v0.51.44)
Total contributors tracked	66	130
Total PR credits	142	568
Top contributors (5+ PRs)	5	20
Sustained (3–4 PRs)	5	11
Two-PR	9	14
Single-PR	47	85

The jump is driven by two things:

1. A ton has happened in 50 tags. Eight batch releases worth of contributor PRs, plus the ongoing autonomous-AI contributor work (@ai-ag2026, 49 PRs), plus @Michaelyklam ramping from 18 to 81 PRs across this stretch.

2. The counting rule explicitly includes closed-but-absorbed PRs now. A PR that was closed on GitHub but whose content shipped via a batch-release squash (with Co-authored-by trailer) or via salvage rewrite (with CHANGELOG attribution) counts the same as a direct merge. This is what Nathan asked for in the task brief: "we often close PRs, but they are actually still being merged. We want to factor those in as contributions from that author as well."

How credit is compiled

1. Pull all closed PRs from the GitHub API — 1,421 of them, both merged_at != null and merged_at == null (closed-unmerged).
2. Walk CHANGELOG.md release-by-release and extract attributions via these patterns:
   • PR #N by @user (canonical bullet form)
   • (#N by @user, (PR #N by @user, (#N, @user;
   • PRs #A, #B by @user (plural)
   • @user — PR #N, @user — N PR (#A, #B)
   • (credit: @user) and (credit: @userA and @userB)
3. For every PR# mentioned in CHANGELOG, union the explicit @-attributed users with the GitHub PR author (when external). Maintainer accounts (@nesquena, @nesquena-hermes) are excluded.
4. For PRs merged on GitHub but not in CHANGELOG (very early PRs, non-noteworthy direct merges), credit the GH author.
5. Three salvaged-design contributors not directly in CHANGELOG are credited in the special-thanks roll: @indigokarasu (Visual redesign: icon rail sidebar + design token system + 7 themes #213 → shaped the v0.50.0 design language), @andrewy-wizard (Add Chinese localization for Hermes WebUI #177 → initial Chinese locale absorbed into v0.42.0), @zenc-cp (fix(hermes): strict ReAct loop — no fabricated tool output #133 → anti-hallucination guard absorbed into streaming.py).

Edge cases handled

• HTML entities like 
 in CHANGELOG were causing false PR# matches. Pre-cleaning step strips them before scan.
• The PR# regex requires a whitespace/paren/bracket preceder so identifiers like 8080:Qwen3-235B, --key=123, and tag-style ##10 don't pollute the count.
• For per-user first/last release computation, merged-on-GH PRs use the smallest tag whose creatordate >= merged_at. Absorbed PRs use the explicit-attribution release (or, when no explicit attribution exists, the earliest CHANGELOG release that mentions the PR#).
• "@user — Closes #N" looks like attribution but the #N is an issue ref, not a PR ref. The GitHub PR author lookup handles this correctly because the PR with the same N (when it exists) gets attributed via path (1).

Diff shape

• CONTRIBUTORS.md: full rewrite of all sections. 61 lines → 93 lines.
• README.md: replaced lines 543-670 (the ## Contributors block, from the top down through ### Bug fix and security contributions) with a top-10 table + 29 notable-contribution blurbs. ~165 lines net.
• No code changes. Docs only.

Sanity checks performed

• Every @user link in CONTRIBUTORS.md verified to point to https://github.com/<same-user> (141 links, 0 mismatches).
• Every notable contributor in README cross-checked to appear in CONTRIBUTORS.md (29/29 present).
• No user mentioned more than twice in CONTRIBUTORS.md.
• kingfish65 was correctly excluded — PR Add slash command system with session management, model switching, and usage analytics #10 was closed-unmerged and never absorbed (
 HTML entity in v0.50.254 was the false-positive match, fixed by the pre-cleaning step).
• deboste's range correctly shows v0.18 era when computed via merged_at tag lookup. The original v0.16.1 claim in the prior CONTRIBUTORS.md was earlier than the actual tag.

If your name is missing or your count looks wrong, open a follow-up PR against this file — the cross-check is automated but the curation matters.