perf(cosmos): optimize partition key range cache memory usage

Three optimizations to reduce CollectionRoutingMap memory footprint:

1. Strip unused fields: Convert 13-field service response dicts to compact
   PKRange namedtuples retaining only id, minInclusive, maxExclusive, parents.
   PKRange supports dict-style access for backward compatibility.

2. Add __slots__ to Range class: Eliminates per-instance __dict__ overhead
   (~100 bytes per Range object).

3. Skip redundant .upper() calls: Service returns uppercase hex strings.
   Avoid creating unnecessary string copies when already uppercase.

Measured with tracemalloc, 150 CosmosClient instances, PPCB enabled:

| Clients | Original | Strip Only | All 3 Patches | PPCB=false |
|---------|----------|------------|---------------|------------|
|      25 |  23.0 MB |    20.5 MB |       20.0 MB |    17.9 MB |
|      50 |  31.9 MB |    27.4 MB |       25.8 MB |    21.7 MB |
|     100 |  44.9 MB |    39.9 MB |       36.6 MB |    29.4 MB |
|     150 |  63.8 MB |    52.9 MB |       43.3 MB |    36.4 MB |

PPCB overhead reduction:

| Clients | Original |  All 3 | Reduction |
|---------|----------|--------|-----------|
|      25 |  5.1 MB  | 2.1 MB |     -58%  |
|      50 | 10.3 MB  | 4.1 MB |     -60%  |
|     100 | 15.4 MB  | 7.2 MB |     -53%  |
|     150 | 27.4 MB  | 6.9 MB |     -74%  |

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: tvaron3

sdk/cosmos/azure-cosmos/azure/cosmos/_routing/aio/routing_map_provider.py

@@ -28,6 +28,7 @@
 from ... import _base
 from ..collection_routing_map import CollectionRoutingMap
 from .. import routing_range
+from ..routing_range import PKRange
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -97,6 +98,18 @@ async def init_collection_routing_map_if_needed(
             # causing the partitionKeyRanges to have both the children ranges and their parents. Therefore, we need
             # to discard the parent ranges to have a valid routing map.
             collection_pk_ranges = list(PartitionKeyRangeCache._discard_parent_ranges(collection_pk_ranges))
+            # Convert to compact PKRange namedtuples to reduce memory footprint
+            collection_pk_ranges = [
+                PKRange(id=r["id"], minInclusive=r["minInclusive"],
+                        maxExclusive=r["maxExclusive"], parents=r.get("parents"))
+                for r in collection_pk_ranges
+            ]
+            # Convert to compact PKRange namedtuples to reduce memory footprint
+            collection_pk_ranges = [
+                PKRange(id=r["id"], minInclusive=r["minInclusive"],
+                        maxExclusive=r["maxExclusive"], parents=r.get("parents"))
+                for r in collection_pk_ranges
+            ]
             collection_routing_map = CollectionRoutingMap.CompleteRoutingMap(
                 [(r, True) for r in collection_pk_ranges], collection_id
             )
@@ -124,7 +137,7 @@ async def get_range_by_partition_key_range_id(
     def _discard_parent_ranges(partitionKeyRanges):
         parentIds = set()
         for r in partitionKeyRanges:
-            if isinstance(r, dict) and routing_range.PartitionKeyRange.Parents in r:
+            if routing_range.PartitionKeyRange.Parents in r:
                 for parentId in r[routing_range.PartitionKeyRange.Parents]:
                     parentIds.add(parentId)
         return (r for r in partitionKeyRanges if r[routing_range.PartitionKeyRange.Id] not in parentIds)

sdk/cosmos/azure-cosmos/azure/cosmos/_routing/routing_map_provider.py

@@ -28,6 +28,7 @@
 from .. import _base
 from .collection_routing_map import CollectionRoutingMap
 from . import routing_range
+from .routing_range import PKRange
 from .routing_range import PartitionKeyRange
 
 _LOGGER = logging.getLogger(__name__)
@@ -82,6 +83,18 @@ def init_collection_routing_map_if_needed(
             # causing the partitionKeyRanges to have both the children ranges and their parents. Therefore, we need
             # to discard the parent ranges to have a valid routing map.
             collection_pk_ranges = list(PartitionKeyRangeCache._discard_parent_ranges(collection_pk_ranges))
+            # Convert to compact PKRange namedtuples to reduce memory footprint
+            collection_pk_ranges = [
+                PKRange(id=r["id"], minInclusive=r["minInclusive"],
+                        maxExclusive=r["maxExclusive"], parents=r.get("parents"))
+                for r in collection_pk_ranges
+            ]
+            # Convert to compact PKRange namedtuples to reduce memory footprint
+            collection_pk_ranges = [
+                PKRange(id=r["id"], minInclusive=r["minInclusive"],
+                        maxExclusive=r["maxExclusive"], parents=r.get("parents"))
+                for r in collection_pk_ranges
+            ]
             collection_routing_map = CollectionRoutingMap.CompleteRoutingMap(
                 [(r, True) for r in collection_pk_ranges], collection_id
             )

sdk/cosmos/azure-cosmos/azure/cosmos/_routing/routing_range.py

@@ -27,6 +27,31 @@
 import json
 
 
+from collections import namedtuple
+
+_PKRangeBase = namedtuple('PKRange', ['id', 'minInclusive', 'maxExclusive', 'parents'])
+
+
+class PKRange(_PKRangeBase):
+    """Compact partition key range with dict-compatible access."""
+    __slots__ = ()
+
+    def __getitem__(self, key):
+        try:
+            return getattr(self, key)
+        except AttributeError:
+            raise KeyError(key)
+
+    def get(self, key, default=None):
+        return getattr(self, key, default)
+
+    def __contains__(self, key):
+        return key in self._fields
+
+    def items(self):
+        return zip(self._fields, self)
+
+
 class PartitionKeyRange(object):
     """Partition Key Range Constants"""
 
@@ -39,6 +64,8 @@ class PartitionKeyRange(object):
 class Range(object):
     """description of class"""
 
+    __slots__ = ('min', 'max', 'isMinInclusive', 'isMaxInclusive')
+
     MinPath = "min"
     MaxPath = "max"
     IsMinInclusivePath = "isMinInclusive"
@@ -50,8 +77,8 @@ def __init__(self, range_min, range_max, isMinInclusive, isMaxInclusive):
         if range_max is None:
             raise ValueError("max is missing")
 
-        self.min = range_min.upper()
-        self.max = range_max.upper()
+        self.min = range_min if range_min == range_min.upper() else range_min.upper()
+        self.max = range_max if range_max == range_max.upper() else range_max.upper()
         self.isMinInclusive = isMinInclusive
         self.isMaxInclusive = isMaxInclusive