Discontiguous PageProtect GC support

[wenyuzhao]

This CL fixes PageProtect's memory unprotect calls. When using discontiguous space,

• If the allocated pages are not in a new chunk, unprotect the pages as before.
• If the pages are in a new chunk, there can be two cases:
  • If the new chunk is not mapped, no need to unprotect
  • If the new chunk is mapped, this means that the chunk is allocated and released on previous GC epochs before. The GC should unprotect these pages.

[qinsoon]

Can you explain a bit why we need this change? What is the bug before this change?

[wenyuzhao]

Can you explain a bit why we need this change? What is the bug before this change?

This is a necessary fix to make page protect gc be able to run with compressed pointers, which uses discontiguous heap

[qinsoon]

The JikesRVM failure should not be related with the PR. We can rerun the test.

[qinsoon]

LGTM