cmd/compile: enable carry chain scheduling for arm64

cmd/compile: enable carry chain scheduling for arm64

This is a follow up of CL 393656 on arm64.

This CL puts ScoreCarryChainTail before ScoreMemory and after
ScoreReadFlags, so that the scheduling of the carry chain will not
break the scheduling of ScoreVarDef.

Benchmarks:
name                                  old time/op    new time/op    delta
ScalarMult/P256-8                       42.0µs ± 0%    42.0µs ± 0%   -0.13%  (p=0.032 n=5+5)
ScalarMult/P224-8                        135µs ± 0%      96µs ± 0%  -29.04%  (p=0.008 n=5+5)
ScalarMult/P384-8                        573µs ± 1%     355µs ± 0%  -38.05%  (p=0.008 n=5+5)
ScalarMult/P521-8                       1.50ms ± 4%    0.77ms ± 0%  -48.78%  (p=0.008 n=5+5)
MarshalUnmarshal/P256/Uncompressed-8     505ns ± 1%     506ns ± 0%     ~     (p=0.460 n=5+5)
MarshalUnmarshal/P256/Compressed-8      6.75µs ± 0%    6.73µs ± 0%   -0.27%  (p=0.016 n=5+5)
MarshalUnmarshal/P224/Uncompressed-8     927ns ± 0%     818ns ± 0%  -11.76%  (p=0.008 n=5+5)
MarshalUnmarshal/P224/Compressed-8       136µs ± 0%      96µs ± 0%  -29.58%  (p=0.008 n=5+5)
MarshalUnmarshal/P384/Uncompressed-8    1.77µs ± 0%    1.36µs ± 1%  -23.14%  (p=0.008 n=5+5)
MarshalUnmarshal/P384/Compressed-8      56.5µs ± 0%    31.9µs ± 0%  -43.59%  (p=0.016 n=5+4)
MarshalUnmarshal/P521/Uncompressed-8    2.91µs ± 0%    2.03µs ± 1%  -30.32%  (p=0.008 n=5+5)
MarshalUnmarshal/P521/Compressed-8       148µs ± 0%      68µs ± 1%  -54.28%  (p=0.008 n=5+5)

Change-Id: I4bf4e3265d7e1ee85765ff2bf006ca5a794d4979
Reviewed-on: https://go-review.googlesource.com/c/go/+/432275
Reviewed-by: Carlos Amedee <carlos@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Eric Fang <eric.fang@arm.com>