Consolidate low-level DP primitives into dpsynth.local_mode.primitives

- Merges _quantiles.py into primitives.py, establishing primitives.py as the
  singular home for low-level DP building blocks in dpsynth.
- Extracts a general-purpose exponential_mechanism primitive which the
  histogram median algorithm now calls.
- Deletes the obsolete DP-SIPS partition selection implementation and tests.
- Updates the module header to document that all primitives expect
  pre-computed aggregates (counts) rather than computing aggregations
  themselves, placing caller responsibility on satisfying privacy assumptions
  (such as each user contributing at most one record to one bucket).
- Updates initialization and tests to import from primitives.py and deletes
  _quantiles.py and _quantiles_test.py.

PiperOrigin-RevId: 976794904

Author: Ryan McKenna

docs/mechanism_api.md

@@ -280,7 +280,7 @@ composed or split as cleanly:
     bisections, whose composite `dp_event` is a composition of exponential
     mechanisms.
 
-[dp-quantiles-src]: https://github.com/google/dpsynth/blob/main/dpsynth/local_mode/_quantiles.py
+[dp-quantiles-src]: https://github.com/google/dpsynth/blob/main/dpsynth/local_mode/primitives.py
 
 Because these heterogeneous parameters cannot be directly combined,
 `configure()` accepts a single scalar `zcdp_rho` and translates it into each

dpsynth/local_mode/_quantiles.py

@@ -22,7 +22,6 @@
 import dp_accounting
 from dpsynth import api
 from dpsynth import domain
-from dpsynth.local_mode import _quantiles
 from dpsynth.local_mode import primitives
 from dpsynth.local_mode import vectorized_transformations as vtx
 import numpy as np
@@ -90,7 +89,7 @@ def compute_grid_spec(
   """Returns (lower, upper, grid_size) for the quantile candidate grid."""
   min_value = float(attribute.min_value)
   if attribute.dtype == 'int':
-    m = _quantiles.jitter_factor(num_partitions)
+    m = primitives.jitter_factor(num_partitions)
     budget = max(2, max_grid_size // m)
     int_range = int(attribute.max_value - attribute.min_value + 1)
     step = max(1, math.ceil(int_range / budget))
@@ -212,7 +211,7 @@ def from_summary(
   ) -> NumericalMeasurement:
     """Returns a NumericalMeasurement from pre-aggregated histogram counts."""
     jitter_strategy = 'refine' if self.attribute.dtype == 'int' else 'symmetric'
-    indices = _quantiles.quantiles_from_histogram(
+    indices = primitives.quantiles_from_histogram(
         rng,
         counts,
         epsilon_levels=np.asarray(self.epsilon_levels),