tcg/aarch64: Reorg goto_tb implementation

The old implementation replaces two insns, swapping between

	b	<dest>
	nop
	br	x30
and
	adrp	x30, <dest>
	addi	x30, x30, lo12:<dest>
	br	x30

There is a race condition in which a thread could be stopped at
the PC of the second insn, and when restarted does not see the
complete address computation and branches to nowhere.

The new implemetation replaces only one insn, swapping between

	b	<dest>
	br	tmp
and
	ldr	tmp, <jmp_addr>
	br	tmp

Reported-by: hev <r@hev.cc>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>