Various fixes to TLS for MIPS.

[jhb]

Summary

• Clear the current thread's TLS pointer on exec. Previously the TLS pointer (and register) remain unchanged.

• Explicitly clear the TLS pointer when new threads are created.

• Make md_tls_tcb_offset per-process instead of per-thread.
  
  The layout of the TLS and TCB are identical for all threads in a process, it is only the TLS pointer values themselves that vary by thread. This also makes setting md_tls_tcb_offset in cpu_set_user_tls() redundant with the setting in exec_setregs(), so only set it in exec_setregs().

Test Plan

• booted mips64 in qemu. ntpd is multi-threaded so was a good test of this (it crashed for some earlier verions when I had a bug)

[jhb]

Note though that qemu doesn't advertise UserLocal support, so my testing didn't verify the changes in cpu_switch() in particular.

[jhb]

@adrian any thoughts?

[adrian]

it /looks/ ok to me; but I can test it on mips32 platforms if you'd like.

Do we have any useful TLS tests that I can run, or should I go write something real quick?

[jhb]

jemalloc uses TLS and ntpd is threaded, so if you boot to multiuser and have ntpd enabled you probably pass. (booting in QEMU with mips64 with ntpd active broke when I had a bug in the threaded case)