Untangle spaghetti in process_directory()
- The code checked for excessive dynamic space size
only *after* mapping dynamic space
- And then checked for mismatched expected/actual address
only *after* mapping at the wrong address.
- And for a bogus space ID only *after* mapping it
- And repeated the lossage message in 6 different places
Other changes:
- always assign anon_dynamic_space_start even if not on linux
- abstract out the HPUX noise
- delay allocation of gencgc page tables until all spaces have been loaded
from the core file so that calloc() won't claim memory that would have been
preferred for dynamic space, which is no longer made in allocate_spaces().