[enhancement] Promote GUIDE_GAP_WEIGHT_THRESHOLD to aiConfig.data for human tuning

Context

PR #10084 (resolves #10035) introduced GUIDE_GAP_WEIGHT_THRESHOLD = 0.8 as a named constant inside ai/daemons/services/GapInferenceEngine.mjs with documented derivation from the ConceptService weight formula. Deferred config-lifting to this follow-up ticket to keep the refactor's scope focused on the architectural change (regex + LLM → graph traversal).

The threshold governs when a concept lacking an EXPLAINED_BY edge emits a [GUIDE_GAP]. Default value 0.8 means "at least tier-1 baseline priority" — every tier-1 concept without a guide qualifies; tier-2/3 concepts qualify only if uniqueness + coverage-deficit bump them above. The derivation is stable, but future ontology growth (post-#10050/#10036/#10037) may reveal the default needs tuning without triggering a code change.

The Fix

1. Promote the constant to aiConfig.data.guideGapWeightThreshold in ai/mcp/server/memory-core/config.template.mjs
2. Preserve the default value (0.8)
3. Preserve the derivation comment — move it from the constant's JSDoc to the config-template comment so curators see the rationale when tuning
4. Import + consume the config value in GapInferenceEngine.inferConceptGraphGaps
5. Run node ./buildScripts/ai/initServerConfigs.mjs will pick up the new config key on install — verify no regeneration needed on existing installs (should be fine since it's additive)

Acceptance Criteria

• aiConfig.data.guideGapWeightThreshold exists in config.template.mjs with default 0.8 + derivation comment
• GapInferenceEngine reads the value from config, not from a file-local constant
• Existing PR #10084 tests still pass (regression guard)
• New test: threshold override via config produces different gap-emission behavior

Origin Session ID

62d6f155-e57f-4279-9b59-36c9e4ecbc5e — flagged as out-of-scope in PR #10084 body

Related

• Parent epic: #10030
• Origin PR: #10084
• Weight formula: ConceptService.calculateWeight — tier_score (0.8/0.5/0.3) + uniqueness (0.2) + coverage_deficit (0.3)