fix(cmd_userinfo): align the +userinfo value column, independent of tab width (#459)

[Aybook]

Closes #459.

The +userinfo value column was ragged (your testhub screenshot: Nick: / Version: / Sent: landing in different columns). Root cause: the labels come from the translatable lang file with inconsistent widths, and the code prepended a fixed tab count to each value (\t\t, inconsistently a single \t for Sent/Received). Tabs snap to 8-column stops, so the value column moved with the label length and with each client's tab-stop width.

Fix

• A small align_labels pass pads every value line's label to one width, computed from the actual labels (so it is correct in any language and re-derives automatically if a label is retranslated), and the fixed tab prefixes are removed. Space-padding in a monospace chat is the only client-independent alignment.
• Banner/blank lines are untouched; the second %s on Level: %s [ %s ] rides along in the tail.
• Lang files are not touched - align_labels normalises the loaded format string (Lua's [[ ]] already folds CRLF to \n), so there is no per-language space-counting to maintain.

Before (v0.23, tabs) vs after (v0.24), rendered:

Nick: 		[HUBOWNER]Aybo          Nick:       [HUBOWNER]Aybo
Version:  	AirDC++ 4.30            Version:    AirDC++ 4.30
Sent:        	686 B               Sent:       686 B
Received: 	41.68 KB                Received:   41.68 KB

Removing the tab prefix also turns two dead X or msg_unknown fallbacks (share, slots) live instead of a latent nil-concat crash - strictly more robust, invisible for normal values.

cmd_userinfo v0.23 -> v0.24. align_labels is exposed as an _-prefixed test seam, so cmd_userinfo joins the seam-only plugins in docs/PLUGIN_API.md §8 (updated four -> five).

Tests (§1a.7)

New tests/unit/cmd_userinfo_align_test.lua, both smoke.yml legs. Drives align_labels against the real msg_userinfo of both shipped lang files and asserts the raw format's value column is ragged (the bug) while the aligned one is uniform, in en and de independently, plus banner/label/%s-count preservation. A source scan guards that no fixed tab prefix returns (locks the other half of the fix). 20/20 green; plugin_lang_test unaffected (lang files untouched).

Review (§1a.6)

Independent reviewer + maintainer spot-check: no blockers, no regressions. Confirmed the alignment invariant holds (all values at column 12), the two regex passes pick the same label boundary, the round-trip is byte-faithful, all 24 tab prefixes removed, and the test is non-vacuous. Both NITs folded in: a source-scan assertion locking the tab-removal, and a comment noting #core is a byte width (labels are ASCII by the keep-DC-terms-English convention).

Cosmetic only, no functional change. 3.2.x only, not backported.

🤖 Generated with Claude Code