Add getcpu() and improve clock_gettime in VDSO

Current VDSO (implemented by sentry, and used by the real application) is a partial implimentation. More specifically, getcpu() is not implemented, and clock_gettime() only works for CLOCK_REALTIME, CLOCK_BOOTTIME, and CLOCK_MONOTONIC.

For example, we need to CLOCK_MONOTONIC_COARSE. Considering that it's just a less precise version of CLOCK_MONOTONIC, let's just use the implementation of CLOCK_MONOTONIC?

getcpu() could be very tricky. Linux uses [RDPID (Read Processor ID) instruction](https://github.com/torvalds/linux/blob/master/arch/x86/include/asm/segment.h#L266). But sentry maintains a set of virtual CPUs.