* switch all .S files to unified asm, and use a new macro `pico_default_asm_setup` to setup compiler to help porting to other compilers. Also some minor tweaks:
* switch some code to use more recent helper methods (e.g. busy_wait_at_least_n_cycles)
* add documentation to host divider header (I had this ages ago and never promoted)
* fixup erroneous docs about 32p32 values in all divider headers
* fix some compiler warnings
* rename recently added `unified_asm` macro to `pico_default_asm`
* Fix various non-GCC warnings (no effect on GCC)
* Reduce use of typeof since non GCC compilers may not support it
* Introduce PICO_C_COMPILER_IS_GNU, PICO_C_COMPILER_IS_CLANG, PICO_C_COMPILER_IS_IAR to CMake as if (CMAKE_C_COMPILER_ID STREQUAL "xxx") is a bit verbose
* Use "unified_asm" macro for all inline asm (it is "volatile __asm" on GNU with a .syntex unified)
* Use NOLOAD instead of COPY in linker scripts (arguably more correct anyway)
* Use the same style for setting _etext in all 4 linker scripts (to the beginning of .data). Clang aligns .data on a 16 byte boundary. Note ideally we'd
add a new symbol __data_source, however that would break backwards compatibility with existing user linker scripts
* Use "a" for .stack, .heap sections because clang complains otherwise, and they are explicitly NOLOAD anyway
* Avoid duplicating __sev, __wfe, __wfi which Clang sometimes seems to provide as built-ins
* Add missing kitchen_sink_blocked_ram binary
* Allow build with LLVM Embedded Toolchain Form ARM v 14.0.0 (unsupported atm)
- The divider state needs to be saved for __aeabi_ddiv, __aeabi_fdiv, __aeabi_dtan and __aeabi_ftan or they won't work in interrupts *(probably not used much youd hope), or on an RTOS context switch
- Refactored code out for the integer and floating point cases
- Improved the floating point 'tests' in passing to check more return values against GCC implementations
- Added floating point usage to the IRQ nesting test case
