Implement Chunked Serial Conformation Map-Reduce

[harryswift01]

Goal

Implement the conformational map-reduce workflow using frame chunks, while still executing serially.

Background

Before adding Dask, the chunked algorithm should be proven correct in serial. This reduces the risk of mixing algorithmic changes with distributed execution issues.

Scope

• Add conformational chunk/task/partial result structures.

• Reuse existing frame chunking helpers where appropriate.

• Reuse ExecutionPolicy where appropriate for internal chunk sizing.

• Implement Pass 1:

  • iterate through selected frame chunks
  • collect chunk-local dihedral angle observations or partial histogram data

• Implement Reduce 1:

  • merge chunk-level angle/histogram partials
  • identify global dihedral peaks/states

• Implement Pass 2:

  • iterate through selected frame chunks
  • assign conformational state labels using global peak definitions
  • compute chunk-local flexible-dihedral counts

• Implement Reduce 2:

  • merge UA conformational state labels
  • merge residue conformational state labels
  • merge flexible-dihedral counts using current semantics

• Preserve deterministic chunk-order reduction.

• Keep execution serial in this sub-issue.

• Do not introduce Dask yet.