Port back changes from open_agb_firm. Now with proper screen init.

Hardware code cleanup + new documentaion/findings.

Don't disable vfp (no harm in leaving it on). Lots of small, mostly style changes.

Added GPU reset code (unused because it works as is). Disabled tag RAM and L1 parity checking on ARM11. Don't rely on code overhead at SD card init for the 74 clock pulses. Implemented new cycle accurate delay function.

More code improvements/documentation.

Improved RSA code a bit. Now uses interrupts.

A few code optimizations.

Updated SPI code. Added unused SPICARD code.

Minor fixes/adjustments to SPI code. Moved spiflash code to ARM11.

Removed AES_subCounter() and replaced AES_addCounter() with optimized version.

Added SHA + NDMA prototype + some notes.

Introducing iomemcpy() and iomemset().

Don't use volatile to prevent reordering.

Corrected OCR register value on SD init to the one from Process9.

Cleanup sdmmc.c/h a bit.

Adjust eMMC clock to be in spec before switching to HS mode.

Optimize SD card poweron for the lowest possible powerup time.

Fixed bugs in sdmmc_send_command(). Partially implemented high speed mode.

[SDMMC] Enable 4 bit mode after switching and always clear fifo.

Disable SD pullup. Set clock back to 16 MHz since the shitty Lexar SDs don't work with 33 MHz.

Update remaining SDMMC clock comments.

Correct comments in sdmmc.c.

Adjust SD init to be in specs.

Remove unneeded reg writes. Set card detect timeout(?) permanently to 0.

Use higher clocks in MMC driver. Needs careful testing!

Increase ARM9 TIMER_sleep() accuracy because why not.

Timers: Use integer math on ARM9 and faster fp math on ARM11.

[ARM9] Cache code optimizations.

Updated NDMA documentation.

Renamed members of _uX types.