Tidy up skipping logic.

[vext01]

This does two things:

1. In many places, when checking if a transform should skip a certain function, we check:

• if the function is marked yk_outline
• if the function is an optimised clone.

Optimised clones are always yk_outline, so the check is redundant there.

2. In NoCallsInEntryBlocks.cpp, this is a transform we don't want to apply to anything yk_outline, but we only skip optimised clones, so we could slow down non-cloned functions marked yk_outline.

(I also tidied up some includes)

Measuring this, I get:

 Benchmark                 Datum1 (ms)  Datum0 (ms)  Ratio  Summary
 cd/yklua/250              37568 ± 867  36190 ± 221   0.96  3.67% faster
 HashIds/yklua/6000        10328 ± 126  10707 ± 171   1.04  3.68% slower
 bounce/yklua/1500         10654 ± 529  10266 ± 243   0.96  indistinguishable
 richards/yklua/100        48768 ± 154  47483 ±1173   0.97  indistinguishable
 towers/yklua/600          11378 ± 203  11088 ± 172   0.97  indistinguishable
 nbody/yklua/250000         5780 ± 400   5653 ±  88   0.98  indistinguishable
 fannkuchredux/yklua/10    14895 ± 991  14698 ±1127   0.99  indistinguishable
 sieve/yklua/3000           4752 ±  77   4713 ±  50   0.99  indistinguishable
 spectralnorm/yklua/1000   14983 ± 186  14876 ±  44   0.99  indistinguishable
 Heightmap/yklua/2000       9500 ± 113   9480 ± 106   1.00  indistinguishable
 fasta/yklua/500000         4093 ±  30   4093 ±  23   1.00  indistinguishable
 bigloop/yklua/1000000000  18226 ±   2  18231 ±   7   1.00  indistinguishable
 json/yklua/100            13841 ±  93  13899 ±  99   1.00  indistinguishable
 mandelbrot/yklua/500       1168 ±   1   1175 ±  10   1.01  indistinguishable
 binarytrees/yklua/15      11798 ± 164  11871 ± 134   1.01  indistinguishable
 list/yklua/1500            8906 ± 166   8963 ± 154   1.01  indistinguishable
 havlak/yklua/1500         64940 ± 311  65520 ±1662   1.01  indistinguishable
 deltablue/yklua/12000     10365 ± 137  10482 ± 140   1.01  indistinguishable
 storage/yklua/1000        28106 ± 191  28430 ± 282   1.01  indistinguishable
 queens/yklua/1000          6502 ±  18   6583 ± 170   1.01  indistinguishable
 knucleotide/yklua/        14782 ± 117  14983 ± 152   1.01  indistinguishable
 revcomp/yklua/             3534 ± 313   3622 ± 111   1.02  indistinguishable
 LuLPeg/yklua/             14927 ± 385  15403 ± 266   1.03  indistinguishable
 permute/yklua/1000         9838 ± 239  10233 ± 548   1.04  indistinguishable

The hashids effect may be real, but offline @ltratt and I agreed that in this case it's probably best to merge this, as the before behaviour is weird and likely to cause confusion.