Add percentile measures (P25, P75, P95, P99)

[stumelius]

Overview

Currently only the median (P50) is computed. Extending this to P25, P75, P95, and P99 would give full box-plot coverage and make outlier detection practical without leaving dbt-profiler.

Motivation

• P25/P75 (IQR) are the standard for identifying outliers and understanding distribution shape
• P95/P99 are essential for SLA/latency-style columns and catching extreme values
• Users doing data quality monitoring almost always want these alongside mean and stddev

Implementation notes

The per-adapter measure_median infrastructure is already in place. Each adapter override would need additional percentile calls:

Adapter	Approach
Default (Postgres, Redshift)	percentile_cont(0.25) within group (order by col)
BigQuery	approx_quantiles(col, 100)[offset(25)]
SQL Server	Extend the existing ROW_NUMBER() approach
Snowflake	percentile_cont(0.25) within group (order by col)
Databricks/Athena	percentile_approx(col, 0.25)

API design

New optional measures (excluded via exclude_measures like existing ones):

{{ dbt_profiler.get_profile(relation=ref('my_model'), exclude_measures=['p25', 'p75', 'p95', 'p99']) }}

Or grouped under a single toggle (e.g. exclude_measures=['percentiles']) to avoid cluttering the interface — worth discussing.