cmd/compile: improve PPC64.rules for loads and stores as found in encoding/binary ByteOrder interface functions

[laboger]

Use PPC64.rules to improve performance of statements that do multiple single byte loads or stores on consecutive bytes, such as is found in the encoding/binary ByteOrder interface. Instead do a single multi-byte load or store where possible, such as lhz, lwz, ld, sth, stw, std.

This improves functions that are found in encoding/binary:
Uint16([]byte) uint16
Uint32([]byte) uint32
Uint64([]byte) uint64
PutUint16([]byte, uint16)
PutUint32([]byte, uint32)
PutUint64([]byte, uint64)

The initial change for this will handle little endian cases.

[gopherbot]

Change https://golang.org/cl/74410 mentions this issue: cmd/compile: add rules to improve consecutive byte loads and stores on ppc64le

[gopherbot]

Change https://golang.org/cl/98136 mentions this issue: cmd/compile: combine byte loads and stores on ppc64le

[bradfitz]

Marking release-blocker (because it has a CL) and @randall77 as owner for review.