Add grouped parallel I/O for graph placements of interleaved reads

[Sagorikanag]

Changelog Entry

• libvgio can now group consecutive alternative placements for interleaved GAM and GAF read pairs while supporting an initial single-threaded processing phase.

Description

This adds grouped interleaved input helpers for callers that need all alternative paired placements of a fragment together.

For GAM and GAF input, consecutive interleaved pairs with the same fragment identity are collected into:

vector<pair<Alignment, Alignment>>

and passed to the caller as one unit. An unmatched final alignment in interleaved input is reported as an error.

The grouped iterator also supports processing groups serially until a caller-provided readiness condition becomes true, after which processing continues in parallel. This allows callers to safely learn shared state, such as a fragment-length distribution, before dispatching groups across worker threads.

Existing unpaired, paired, and grouped APIs retain their previous behavior through compatibility wrappers.