Fix user scoring bugs — inverted spam_label, wrong quotient scale

[realproject7]

Problem

The user scoring script (`scripts/score-users.mjs`) from PR #659 has critical logic errors that produced wrong results for the majority of users.

Bugs to fix

Bug 1: `spam_label` logic is INVERTED (CRITICAL)

Current (wrong):
```javascript
const isSpam = spamRaw.length > 0 && spamRaw !== "0";
if (isSpam) return { score: 0, tag: "None" };
```

Reality:

• `spam_label = 2` → non-spammer (good user) — 4,848 users
• `spam_label = 0` → not verified / potential spam — 2,435 users
• `spam_label = ""` (blank) → unknown — 52 users

Fix:
```javascript
// spam_label=2 means non-spammer. 0 or blank = not verified.
// Don't disqualify based on spam_label — use as a scoring signal instead.
const spamVerified = spamRaw === "2"; // verified non-spammer
// Add bonus for verified non-spammer, slight penalty for unverified
```

Impact: 4,848 legitimate users (including project7 with 8,978 followers) were incorrectly scored 0.

Bug 2: `quotient_score` scale is wrong

Current (wrong):
```javascript
const quotientScore = Math.min(1, num("quotient_score") / 100);
```

Reality: Quotient score range is 0 to ~0.95 (already 0-1 scale). Dividing by 100 makes it effectively zero.

Fix:
```javascript
const quotientScore = Math.min(1, Math.max(0, num("quotient_score")));
```

Impact: Quotient signal contributed ~0% to scores for all 5,775 users with quotient data.

Re-run requirements

After fixing the logic:

1. Re-run the script: `node scripts/score-users.mjs`
2. Output to same file: `archive/dropcast-users-scored-20260330.csv`
3. Post updated summary stats as ticket comment (tag counts, score distribution, top 20)
4. Verify project7 (FID 8106) scores appropriately high (8,978 followers, 0.99 neynar, 0.706 quotient, X verified, pro subscriber)

Verification reference

project7 expected signals:

• FC followers: 8,978 → log-scaled to 100K = ~0.79
• X followers: 2,389 → log-scaled to 500K = ~0.59
• X verified: true → 1.0
• Neynar score: 0.99
• Quotient score: 0.706
• Pro subscriber: true
• spam_label: 2 (verified non-spammer)
• Bio: 100+ chars, contains "designer", "Coder"
• Has Twitter: yes

Expected score should be 60+, not 0.

Branch

`task/673-fix-user-scoring`

Self-Verification (T3)

• project7 (FID 8106) scores 60+ with tag "Writer"
• spam_label=2 users are NOT zeroed out
• spam_label=0 users get a penalty (not instant zero)
• Quotient scores used as-is (0-1), not divided by 100
• Updated summary stats posted as comment
• Score distribution looks reasonable (top users 70-90 range)
• `npm run build` passes

[realproject7]

Updated Scoring Results

Fixes Applied

1. spam_label=2 (verified non-spammer): no longer zeroed out — only spam_label=1 disqualifies
2. spam_label=0/blank: -3 penalty instead of instant zero; spam_label=2 gets +2 bonus
3. quotient_score: used as-is (0-1 scale), no longer divided by 100

Tag Counts

Tag	Count
Writer	483
Reader	5,757
None	1,095

Score Distribution

Range	Count
80-100	0
60-79	89
40-59	1,208
20-39	2,779
0-19	3,259

Top 20 Users

Rank	Username	Score	Tag	Followers
1	afrochicks	73	Writer	18,024
2	push-	71	Writer	7,790
3	nomadicframe	70	Writer	10,448
4	odysseyheart	69	Writer	2,951
5	to	69	Writer	11,816
6	femmie	69	Writer	13,029
7	kokocodes	69	Writer	6,897
8	bleu.eth	69	Writer	12,771
9	thosmur	68	Writer	5,335
10	anemale.eth	68	Writer	7,083
11	tonyminh	68	Writer	9,306
12	swishh.eth	67	Writer	4,930
13	zaal	67	Writer	2,691
14	bizarrebeast	67	Writer	4,353
15	statuette	67	Reader	11,255
16	nutscity13	66	Writer	2,226
17	trillobyte	66	Writer	4,827
18	fffflood	65	Writer	6,709
19	vmfcoin	65	Writer	2,253
20	greyseymour	65	Writer	7,348

Verification

• project7 (FID 8106): Score 64, Tag Writer ✅