Weighted Score: shared multi-stat sort across PoB - #9845

19 tests across 3 describe blocks:
- WeightedScore module: defaultWeights, getWeights, computeRatioScore
  (neutrality, ranking, edge cases: inf/zero/missing, FullDPS fallback)
- TradeQueryGenerator delegation: result matches direct call, ranking preserved
- Tree integration: stat registered, power builder completes, powerMax >= 0

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

NotableDBControl was missing the isWeightedScore propagation in
BuildSortOrder and the computeRatioScore branch in ListBuilder,
causing all notables to score 0 when sorted by Weighted Score.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Consumers of powerStatList that need to compute a stat value from a
calc output (e.g. RadiusJewelFinder's getImpactValue) cannot read
output["WeightedScore"] directly since it is not a real calc field.

Provide a getValue(output, build) callback on the stat entry so any
consumer can get a meaningful weighted ratio score without needing to
know about the WeightedScore module.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

LoadModule does not cache — calling it inside getValue on every
invocation created a new module table each time, causing memory
pressure in tight loops like RadiusJewelFinder compute.

Move the LoadModule call to module scope so it executes once at
startup.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

When 'Weighted Score' was selected as Fallback Weight Mode, the timeless
jewel finder's generateFallbackWeights read output["WeightedScore"] — a
field that does not exist in calc output. This returned 0 for every node,
causing all fallback weights to be computed as -100 (or -50 for nodes
with a non-unit divisor), making the weighted score sort meaningless.

Fix: introduce a getStatValue helper inside generateFallbackWeights that
delegates to selection.getValue(rawOutput, build) when present, mirroring
the same pattern used in RadiusJewelFinder:getImpactValue. The raw
(non-Minion-scoped) output is passed so getValue receives the full calc
output as expected by WeightedScore.computeRatioScore.

Also guard baseValue == 0 to avoid division by zero for builds with no
relevant output stat.

Add two tests to TestWeightedScore_spec.lua covering getValue correctness
on the powerStatList entry.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

The fallbackWeightsList dropdown entries were built by copying only stat,
transform, and label from data.powerStatList — getValue was silently
dropped. As a result, generateFallbackWeights received selection.getValue
= nil and could never delegate to the WeightedScore callback, falling back
to reading output["WeightedScore"] (a non-existent field) and producing
weight = -100 for every node.

Fix: copy getValue when building each fallbackWeightsList entry so the
callback reaches generateFallbackWeights correctly.

Add a test asserting that the constructed entry carries getValue.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Route WeightedScore power reports through FullDPS when active weights require it, reuse the cached base output for WeightedScore getValue, and keep allocated nodes out of the default unallocated report list.

…action entry

Replace per-surface Edit Weights buttons (TreeTab heatmap, ItemDB unique
sort) with a shared action entry appended to any sort/heatmap dropdown
that exposes Weighted Score. NotableDB anoint sort gains the same
affordance (no Edit Weights button before).

Why: the ItemDB pane is 360 px wide; the previous Edit Weights button
shared the same anchor as the League dropdown and forced a mutually
exclusive show/hide between League filter and Edit Weights. A user who
wanted to filter by League and sort by Weighted Score could not do both.

WeightedScore.appendEditWeightsAction(list, openEditor) sentinel-checks
for any entry with isWeightedScore and appends a single action entry.
Each consumer dropdown selFunc gates on value.isAction: invokes
value.action() then restores the prior selection via SelByValue.

Tests: 2 new specs in TestWeightedScore_spec (no-op without WS, append
plus invocable callback when WS present). Peer-reviewed by Codex
(approved, no blocking findings).

Restore the shared accessor, transforms, and FullDPS fallback expected by Trade Query after the origin/dev rebase.

Avoid generating a Minion Weighted Score entry and hide the score from modifier sort menus that only read output fields.

Keep negative scores and request FullDPS only when the active weights need it.

Calculate Weighted Score candidates with the matching Full DPS context and retain the baseline output for Tree fallback scoring.

Evaluate contextual metrics against the displayed item baseline so Weighted Score and Full DPS rank candidates correctly.

Evaluate crafted prefix and suffix candidates through the contextual stat API so Weighted Score uses its baseline and Full DPS requirements instead of tying every modifier at zero.

Keep crafted affix, Compare Power, and Timeless fallback scoring aligned with saved weights.