Port ResolveAxesTypes, IsMixedMode, MixedAxes and CSV/JSON mixed-mode
parsers from feat/mixed-axis-scatter. When IsSelectAxisMode, use
SelectViews[0] for axis placement; wire assembleDataset to emit mixed or
value axes and auto-enable 3D for numeric xyz.

Fix IsExplicitGrouping so empty GroupPattern does not block select axis mode.

Parse each solo --select view from a single csv/json read, assemble one
Dataset per view with auto-names and per-view 3D auto-enable, and embed
an array in HTML/JSON output when multiple views are present.

Wire __mixed_mode__ through the worker/pipeline for category X + value
Y/Z scatter datasets. Add buildMixedModeChart, 2D/3D scatter option
builders, settings field filtering, and regenerated embedded UI bundle.

Add design spec superseding mixed-axis --axes approach. Update bar, line,
and scatter chart docs with three-role --select structure. Extend grouping
guide with solo select axis mode and decision table. Add mode matrix to
root command reference. Include region-metrics.csv example for mixed scatter.

- Enable bar/line (2D and 3D) for solo mixed-axis --select views
- Add trailing (Title) syntax for multi-stat chart-tab names; keep {} for axis labels
- Default multi-stat stat names to metric by dim instead of dim & metric
- Merge stats per row when repeatable --select shares the same dimension column
- Propagate dimension {label} to axes[].label in multi-stat datasets
- Document updated --select semantics in root, grouping, and scatter guides

Parallel act matrix jobs with --bind shared one data-input.txt and
corrupted curl JSON before --json-path could parse it.

This reverts commit e3b4843.

…duplication

… value boundary

DispatchSelectMode propagates AxisType to the parser's local cfg copy,
but ParseFunc passes Config by value so the mutation doesn't reach
prepareData/assembleDataset. DatasetAxesForSelectView now falls back to
inferring mixed-ness from DataPoint XAxis values when AxisType is empty,
restoring the pre-refactor behavior. Added end-to-end integration test
(ParseCSV → assembleDataset) as regression guard.

Drop accidentally committed .superpowers and docs/superpowers files
and extend .gitignore so they stay untracked.

Add follow-up tests for pipeline buildDataset branches, flagbag reset/seed,
tabular edge cases, and csv/json kindFn error paths to push patch coverage
above the Codecov threshold.

Cover repeatable --select ParseConfig branches, grouped duplicate detection,
symbol fatal validation, and group slice soft-rule handling.