Change when address environment is instantiated; there are/were locations where access is made to the allocation address environment when it is not yet in place

addrenv interface changes: up_addrenv_create() may need to create .text and .bss/.data separately because of differing access privileges (read/execute vs read/write).  And, as a consequence, up_addrenv_vaddr() needs to be split into up_addrenv_vtext(0 and up_addrenv_vdata().

Add addrenv.h; First cut at Cortex-A address environment structures; Add configuration options to setup address enviornment

Change CONFIG_ADDRENV to CONFIG_ARCH_ADDRENV; change how it is selected -- the architecure must first declare support

include/nuttx/pgalloc.h and mm/mm_pgalloc.c:  Add a simple page allocator based on the existing NuttX granule allocator.  I am not certain if the granule allocator is sufficiently deterministic for long range use, but it gets get a page allocator in place for testing very quickly.

Recent STM32 UART change:  Wasn't that logic backward?  Shouldn't that have been disable the USART if (1) we don't have than many USARTs OR (2) we don't have that particular USART -- not AND.