[kernel] Reduce precision time routines single routine

[ghaerr]

Pursuant to continued discussion in Mellvik/TLVC#71 (comment).

The following changes are made to the precision timing routines:

• get_time_10ms and get_time routines removed.
• get_time_50ms routine renamed get_ptime
• get_ptime returns continuous 0.838us precision from 0.838us to ~42 seconds.
• get_ptime returns unsigned long to allow for greater elapsed time. However, if elapsed time is known to be < 50ms, the returned value can be cast to unsigned int for use without the trouble of subsequent 32-bit handling.
• get_ptime returns 0 on overflow: elapsed time > 42.85 seconds.
• Sub-10ms handling within get_ptime is very quick and only uses unsigned int and no multiplies/divide.
• Corrected bug on jiffies overflow on elapsed time over ~32 seconds.
• Noticed that 5 through 9 seconds is correctly displayed as 5000.000ms through 9999.999ms. Currently not fixed as would add quite a bit of code in printk formatting routine. 1-4 seconds and 10+ seconds displayed properly.

Tested on QEMU and timings are consistent.

Next step will be to start working on user mode C library version of get_ptime which will share almost the same code, except for the method of reading kernel jiffies.