Fix potential parallelism issue in MapToContigs for single-end reads

[YYQ0713]

In MapToContigs, the OpenMP directive for single-end reads lacks a for clause:

#pragma omp parallel reduction(+ : num_added, num_mapped)
for (size_t i = 0; i < lib.seq_count(); ++i) { ... }

This may cause the loop iterations to not be correctly distributed across threads. Adding for ensures proper parallel iteration:

#pragma omp parallel for reduction(+ : num_added, num_mapped)
for (size_t i = 0; i < lib.seq_count(); ++i) { ... }

This change aligns the behavior with the intended parallel processing. I’m not certain if the original omission had a specific reason, so this PR just suggests the adjustment for more consistent parallelism.

[YYQ0713]

#377 #385 I suspect this might be the cause of these problems.