feat(POP-4266): introduce anon stats sampling_rate for di 2d stats

[leonanos8]

Description

Adds sampling support to the shared DI 2D anon-stats pipeline, allowing randomly sampled pairs and fully retained exception pairs to be stored, processed, and published separately.

Details:

• Adds sampling_rate metadata to PostgreSQL rows and published 2D histograms.
• Defines sampling_rate=20 for the unbiased 20% random sample.
• Defines sampling_rate=100 for unsampled pairs retained by the score or opposite-mirror exception rules.
• Separates PostgreSQL counting, fetching, processing, and cleanup by sampling rate and match flags.
• Adds configuration for the sampling rates processed by the anon-stats server.
• Adds a separate minimum job size for sparse fully retained rows.
• Preserves existing behavior by defaulting existing rows and configurations to sampling_rate=100.
• Adds validation and tests for sampling-rate configuration, serialization, and job-size limits.

Notes

@dkales @marsenis @evagelia I would like your perspective here on the isolated separate minimum job size for sparse fully retained rows. This controls on how often we publish histograms for cases that had big scores (>500). Default value is set to 20 but we maybe want to increase this to avoid a narrow time window privacy attack.