runtime: keep pointer bits set always in 1-word spans

runtime: keep pointer bits set always in 1-word spans

It's dumb to clear them in initSpan, set them in heapBitsSetType,
clear them in heapBitsSweepSpan, set them again in heapBitsSetType,
clear them again in heapBitsSweepSpan, and so on.

Set them in initSpan and be done with it (until the span is reused
for objects of a different size).

This avoids an atomic operation in a common case (one-word allocation).
Suggested by rlh.

name                   old mean              new mean              delta
BinaryTree17            5.87s × (0.97,1.03)   5.93s × (0.98,1.04)              ~    (p=0.056)
Fannkuch11              4.34s × (1.00,1.01)   4.41s × (1.00,1.00)            +1.42% (p=0.000)
FmtFprintfEmpty        86.1ns × (0.98,1.03)  88.9ns × (0.95,1.14)              ~    (p=0.066)
FmtFprintfString        292ns × (0.97,1.04)   284ns × (0.98,1.03)            -2.64% (p=0.000)
FmtFprintfInt           271ns × (0.98,1.06)   274ns × (0.98,1.05)              ~    (p=0.148)
FmtFprintfIntInt        478ns × (0.98,1.05)   487ns × (0.98,1.03)            +1.85% (p=0.004)
FmtFprintfPrefixedInt   397ns × (0.98,1.05)   394ns × (0.98,1.02)              ~    (p=0.184)
FmtFprintfFloat         553ns × (0.99,1.02)   543ns × (0.99,1.01)            -1.71% (p=0.000)
FmtManyArgs            1.90µs × (0.98,1.05)  1.88µs × (0.99,1.01)            -0.97% (p=0.037)
GobDecode              15.1ms × (0.99,1.01)  15.3ms × (0.99,1.01)            +0.78% (p=0.001)
GobEncode              11.7ms × (0.98,1.05)  11.6ms × (0.99,1.02)            -1.39% (p=0.009)
Gzip                    646ms × (1.00,1.01)   647ms × (1.00,1.01)              ~    (p=0.120)
Gunzip                  142ms × (1.00,1.00)   142ms × (1.00,1.00)              ~    (p=0.068)
HTTPClientServer       89.7µs × (0.99,1.01)  90.1µs × (0.98,1.03)              ~    (p=0.224)
JSONEncode             31.3ms × (0.99,1.01)  31.2ms × (0.99,1.02)              ~    (p=0.149)
JSONDecode              113ms × (0.99,1.01)   111ms × (0.99,1.01)            -1.25% (p=0.000)
Mandelbrot200          6.01ms × (1.00,1.00)  6.01ms × (1.00,1.00)            +0.09% (p=0.015)
GoParse                6.63ms × (0.98,1.03)  6.55ms × (0.99,1.02)            -1.10% (p=0.006)
RegexpMatchEasy0_32     161ns × (1.00,1.00)   161ns × (1.00,1.00)  (sample has zero variance)
RegexpMatchEasy0_1K     539ns × (0.99,1.01)   563ns × (0.99,1.01)            +4.51% (p=0.000)
RegexpMatchEasy1_32     140ns × (0.99,1.01)   141ns × (0.99,1.01)            +1.34% (p=0.000)
RegexpMatchEasy1_1K     886ns × (1.00,1.01)   888ns × (1.00,1.00)            +0.20% (p=0.003)
RegexpMatchMedium_32    252ns × (1.00,1.02)   255ns × (0.99,1.01)            +1.32% (p=0.000)
RegexpMatchMedium_1K   72.7µs × (1.00,1.00)  72.6µs × (1.00,1.00)              ~    (p=0.296)
RegexpMatchHard_32     3.84µs × (1.00,1.01)  3.84µs × (1.00,1.00)              ~    (p=0.339)
RegexpMatchHard_1K      117µs × (1.00,1.01)   117µs × (1.00,1.00)            -0.28% (p=0.022)
Revcomp                 914ms × (0.99,1.01)   909ms × (0.99,1.01)            -0.49% (p=0.031)
Template                128ms × (0.99,1.01)   127ms × (0.99,1.01)            -1.10% (p=0.000)
TimeParse               628ns × (0.99,1.01)   639ns × (0.99,1.01)            +1.69% (p=0.000)
TimeFormat              660ns × (0.99,1.01)   662ns × (0.99,1.02)              ~    (p=0.287)

Change-Id: I3127b0ab89708267c74aa7d0eae1db1a1bcdfda5
Reviewed-on: https://go-review.googlesource.com/9884
Reviewed-by: Austin Clements <austin@google.com>