parallelize sweeping of object pools

[d-netto]

We've backported this PR and ran some benchmarks internally. It seems to be a win for reducing maximum GC pauses for us.

For the binary_tree open-source GCBenchmarks, concurrent page sweeping seems to be a lot more effective, but this PR doesn't seem to hurt performance there compared to master.

It seems to decrease sweep times significantly in the objarray.jl GCBenchmark:

• master (36 mutator threads, 36 GC threads):

bench = "objarray.jl"
┌─────────┬────────────┬─────────┬───────────┬────────────┬──────────────┬───────────────────┬──────────┬────────────┐
│         │ total time │ gc time │ mark time │ sweep time │ max GC pause │ time to safepoint │ max heap │ percent gc │
│         │         ms │      ms │        ms │         ms │           ms │                us │       MB │          % │
├─────────┼────────────┼─────────┼───────────┼────────────┼──────────────┼───────────────────┼──────────┼────────────┤
│ minimum │      15899 │    9316 │      2915 │       6401 │         2386 │            308397 │     9305 │         57 │
│  median │      17527 │   10430 │      3336 │       7135 │         2937 │            605870 │     9766 │         59 │
│ maximum │      19342 │   12467 │      4317 │       8149 │         3742 │            679771 │    11402 │         64 │
│   stdev │       1036 │     947 │       426 │        538 │          396 │            153045 │      728 │          2 │
└─────────┴────────────┴─────────┴───────────┴────────────┴──────────────┴───────────────────┴──────────┴────────────┘

• PR (36 mutator threads, 36 GC threads):

bench = "objarray.jl"
┌─────────┬────────────┬─────────┬───────────┬────────────┬──────────────┬───────────────────┬──────────┬────────────┐
│         │ total time │ gc time │ mark time │ sweep time │ max GC pause │ time to safepoint │ max heap │ percent gc │
│         │         ms │      ms │        ms │         ms │           ms │                us │       MB │          % │
├─────────┼────────────┼─────────┼───────────┼────────────┼──────────────┼───────────────────┼──────────┼────────────┤
│ minimum │      13273 │    6318 │      3758 │       2557 │         1631 │            282552 │     8704 │         46 │
│  median │      14326 │    6809 │      4139 │       2702 │         1851 │            664444 │    10061 │         48 │
│ maximum │      14865 │    7065 │      4304 │       2771 │         1982 │            760770 │    11342 │         50 │
│   stdev │        547 │     259 │       180 │         86 │          121 │            134561 │      780 │          1 │
└─────────┴────────────┴─────────┴───────────┴────────────┴──────────────┴───────────────────┴──────────┴────────────┘