Add pluggable backend architecture to cuvs-bench (rapidsai#1536)

## Summary

This PR introduces a pluggable backend architecture for `cuvs-bench`, enabling extensibility to support multiple benchmark backend types beyond the current C++ Google Benchmark executables.

## Motivation

Currently, `cuvs-bench` is tightly coupled to C++ Google Benchmark executables via subprocess calls. This limits the ability to:
- Benchmark Python-native libraries (e.g., hnswlib-python, faiss-python)
- Test network-based vector databases (e.g., Milvus, Qdrant, Elasticsearch)
- Allow users to add custom backends without modifying core cuvs-bench code

This PR lays the foundation for a plugin system that addresses these limitations.

## Changes

### Phase 1: Core Infrastructure
- **`backends/base.py`**: Defines the `BenchmarkBackend` abstract interface and data structures (`Dataset`, `BuildResult`, `SearchResult`)
- **`backends/registry.py`**: Implements `BackendRegistry` for centralized plugin management with support for:
  - Dynamic loading from Python modules
  - Configuration-based loading from YAML files
  - Runtime plugin registration
- **`backends/__init__.py`**: Package initialization with auto-registration of built-in backends

### Phase 2: C++ Backend Wrapper
- **`backends/cpp_gbench.py`**: `CppGoogleBenchmarkBackend` wraps existing C++ executables to maintain **100% backward compatibility**
- Handles subprocess execution, temporary file management, and Google Benchmark JSON parsing
- Auto-registered on import as `"cpp_gbench"`

### Testing
- **`tests/test_registry.py`**: Comprehensive unit tests for registry and base infrastructure (20 tests)
- **`tests/test_cpp_gbench.py`**: Unit tests for C++ backend wrapper (8 tests)
- **All 28 tests passing** ✅

## Backward Compatibility

✅ Existing C++ benchmarks continue to work without modification  
✅ No changes required to existing benchmark configurations  
✅ Current CLI and orchestration remain functional

## Future Work

This PR enables future phases:
- **Phase 3**: Python in-process backend (hnswlib-python, numpy-based algorithms)
- **Phase 4**: Network VDB backend (Milvus, Qdrant, Elasticsearch)
- **Phase 5**: Migration of existing `cuvs_bench.run` orchestration to use the plugin system

## Testing Instructions

```bash
cd python/cuvs_bench
python -m pytest cuvs_bench/tests/test_registry.py cuvs_bench/tests/test_cpp_gbench.py -v
```

Closes rapidsai#1100

Authors:
  - Joseph Nke (https://github.com/jnke2016)
  - Corey J. Nolet (https://github.com/cjnolet)

Approvers:
  - Bradley Dice (https://github.com/bdice)
  - Dante Gama Dessavre (https://github.com/dantegd)

URL: rapidsai#1536

Author: jnke2016

conda/environments/all_cuda-129_arch-aarch64.yaml

@@ -46,7 +46,7 @@ dependencies:
 - recommonmark
 - rust
 - scikit-build-core>=0.11.0
-- scikit-learn
+- scikit-learn>=1.5
 - sphinx-copybutton
 - sphinx-markdown-tables
 - sphinx>=8.0.0

conda/environments/all_cuda-129_arch-x86_64.yaml

@@ -46,7 +46,7 @@ dependencies:
 - recommonmark
 - rust
 - scikit-build-core>=0.11.0
-- scikit-learn
+- scikit-learn>=1.5
 - sphinx-copybutton
 - sphinx-markdown-tables
 - sphinx>=8.0.0

conda/environments/all_cuda-131_arch-aarch64.yaml

@@ -47,7 +47,7 @@ dependencies:
 - recommonmark
 - rust
 - scikit-build-core>=0.11.0
-- scikit-learn
+- scikit-learn>=1.5
 - sphinx-copybutton
 - sphinx-markdown-tables
 - sphinx>=8.0.0

conda/environments/all_cuda-131_arch-x86_64.yaml

@@ -47,7 +47,7 @@ dependencies:
 - recommonmark
 - rust
 - scikit-build-core>=0.11.0
-- scikit-learn
+- scikit-learn>=1.5
 - sphinx-copybutton
 - sphinx-markdown-tables
 - sphinx>=8.0.0

conda/environments/bench_ann_cuda-129_arch-aarch64.yaml

@@ -41,7 +41,7 @@ dependencies:
 - pyyaml
 - rapids-build-backend>=0.4.0,<0.5.0
 - requests
-- scikit-learn
+- scikit-learn>=1.5
 - setuptools>=77.0.0
 - sysroot_linux-aarch64==2.28
 - wheel

conda/environments/bench_ann_cuda-129_arch-x86_64.yaml

@@ -44,7 +44,7 @@ dependencies:
 - pyyaml
 - rapids-build-backend>=0.4.0,<0.5.0
 - requests
-- scikit-learn
+- scikit-learn>=1.5
 - setuptools>=77.0.0
 - sysroot_linux-64==2.28
 - wheel

conda/environments/bench_ann_cuda-131_arch-aarch64.yaml

@@ -42,7 +42,7 @@ dependencies:
 - pyyaml
 - rapids-build-backend>=0.4.0,<0.5.0
 - requests
-- scikit-learn
+- scikit-learn>=1.5
 - setuptools>=77.0.0
 - sysroot_linux-aarch64==2.28
 - wheel

conda/environments/bench_ann_cuda-131_arch-x86_64.yaml

@@ -45,7 +45,7 @@ dependencies:
 - pyyaml
 - rapids-build-backend>=0.4.0,<0.5.0
 - requests
-- scikit-learn
+- scikit-learn>=1.5
 - setuptools>=77.0.0
 - sysroot_linux-64==2.28
 - wheel

dependencies.yaml

@@ -450,7 +450,7 @@ dependencies:
     common:
       - output_types: [conda, requirements, pyproject]
         packages:
-          - scikit-learn
+          - scikit-learn>=1.5
   java:
     common:
       - output_types: conda
@@ -488,7 +488,7 @@ dependencies:
           - pandas
           - pyyaml
           - requests
-          - scikit-learn
+          - scikit-learn>=1.5
       - output_types: [conda]
         packages:
           - matplotlib-base>=3.9

python/cuvs/pyproject.toml

@@ -38,7 +38,7 @@ test = [
     "cupy-cuda13x>=13.6.0",
     "pytest-cov",
     "pytest<9.0.0",
-    "scikit-learn",
+    "scikit-learn>=1.5",
 ] # This list was generated by `rapids-dependency-file-generator`. To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
 
 [project.urls]