feat: target treatment lift at a single metric when set

[mohossam01]

What this PR does

Adds per-metric treatment targeting. A treatment cohort can now direct its lift at a single named metric, leaving all other metrics unaffected. When no target metric is set, treatment applies trajectory-wide as before with byte-identical output. The validator rejects target metric names that don't match a declared metric, and the manifest records the target metric per entity and per cohort.

Files Modified

plotsim/config.py
plotsim/validation.py
plotsim/metrics.py
plotsim/tables.py
plotsim/builder/input.py
plotsim/builder/interpreter.py
plotsim/manifest.py
plotsim-schema.json
tests/test_per_metric_treatment.py
tests/test_manifest.py
tests/test_heteroscedastic_noise.py
tests/test_multi_source.py
tests/test_time_varying_correlations.py
CHANGELOG.md
docs/site/manifest-reference.md
docs/site/user-guide/experiments-and-cohorts.md

How to test

python -m pytest tests/test_per_metric_treatment.py
python -m pytest tests/test_treatment_control.py
python -m pytest