feat: dependency-driven build scheduler (no per-region barrier)

[krystophny]

Summary

• Replace the per-region OpenMP barrier in build_package with a
  dependency-driven ready-queue scheduler
• Each target builds as soon as its own dependencies are done, with no barrier
• Workers pop ready targets, build them, and release any dependent whose
  remaining dependency count reaches zero

Context

Today the backend groups targets into schedule regions by dependency depth and
builds each region with !$omp parallel do, so there is an implicit barrier
between regions: region i+1 cannot start until every target in region i
finishes. One slow target (a straggler) stalls the whole next region.

Change

• build_dependency_graph computes, over the queued targets, each target's
  unmet-dependency count and a compressed-sparse-row reverse adjacency
  (dependents of each target)
• The build loop runs one !$omp parallel region: pop a ready target under
  critical(fpm_sched), build it, then release its dependents and bump the
  completion count
• Acyclicity is already guaranteed upstream by sort_target; each target is
  pushed and built exactly once; on failure the queue drains and still releases
  dependents, so the loop always terminates

Merge order

Independent PRs (any order):

• Native is_dir (perf: use native directory checks on Unix #1291)
• Native mkdir (perf: create Unix directories without shell #1292)
• Build only selected run targets (perf: build only selected run targets #1293)
• Skip dry-run for run/test --list (perf: skip dry-run build for run --list and test --list #1294)
• Skip dry-run for build --list (perf: skip dry-run build for build --list #1295)
• Serialize pretty-mode progress under OpenMP (fix: serialize pretty-mode build progress output under OpenMP #1296)
• Dependency-driven build scheduler (feat: dependency-driven build scheduler (no per-region barrier) #1297)

Stacked chain (merge in order):

1. Fortran compiles via argv (feat: exec Fortran compiles via argv (posix_spawn/CreateProcess) #1299)
2. Parallel link/archive via argv (feat: run link and archive in parallel via argv spawn #1300)
3. C/C++ compiles via argv (feat: exec C/C++ compiles via argv, close the shell-fork gap #1301)

Fork-safety fix:

• Serialize mkdir on run_command lock (fix: serialize mkdir shell fork on the run_command lock #1298)

After the above:

• Interface-aware rebuild (Skip recompiling dependents when a module interface is unchanged (gfortran) #1289, already open)
• Enable default features via manifest (pending CI fix)

Verification

$ fpm test --flag -fopenmp
TALLY;TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT
PASSED: all 35 tests passed

• A generated 200-module project builds successfully and deterministically (3/3
  runs), including with OMP_NUM_THREADS=1
• A 24-module dependency chain (width-1 graph, worst case for the scheduler's
  busy-wait) builds correctly
• Not slower than the barrier: bigmod clean build 3.25s vs 3.29s; chain 0.42s
  vs 0.42s (best of 5, 3 reps)

The intended win is on unbalanced graphs where a straggler stalls a region
boundary; the balanced bigmod and the pure chain only show parity (as expected).