refactor: sklearn-style class API for clustering methods

Each clustering algorithm is now a frozen dataclass configured at
construction time and invoked as a callable:

    result = Gromos(cutoff_nm=0.2)(rmsd_matrix)
    result = DBSCAN(eps=0.15, backend="numba")(rmsd_matrix)
    result = KMeans(n_clusters=10)(pca.projections)

Distance-matrix methods: Gromos, Hierarchical, DBSCAN, HDBSCAN
Feature-vector methods: KMeans, MiniBatchKMeans, RegularSpace

DBSCAN supports backend="numba" (default, O(n) aux) and
backend="sklearn" (official). Other methods delegate to their
official library (scipy/sklearn/deeptime).

cluster_conformations and cluster_features remain as backward-
compatible wrappers accepting both string names and class instances.
Removed all memory guards per user request.

Author: FridrichMethod

src/mdpp/analysis/__init__.py

@@ -1,6 +1,13 @@
 """Analysis subpackage for molecular dynamics trajectories."""
 
 from mdpp.analysis.clustering import (
+    DBSCAN,
+    HDBSCAN,
+    Gromos,
+    Hierarchical,
+    KMeans,
+    MiniBatchKMeans,
+    RegularSpace,
     cluster_conformations,
     cluster_features,
     compute_rmsd_matrix,
@@ -31,6 +38,13 @@
 )
 
 __all__ = [
+    "DBSCAN",
+    "HDBSCAN",
+    "Gromos",
+    "Hierarchical",
+    "KMeans",
+    "MiniBatchKMeans",
+    "RegularSpace",
     "cluster_conformations",
     "cluster_features",
     "compute_contact_frequency",