minimize inlining cost and code complexity

OLD:
can inline popcntMaskSlice with cost 32
can inline popcntAndSlice with cost 32
can inline popcntOrSlice with cost 32
can inline popcntXorSlice with cost 32

github.com/bits-and-blooms/bitset.popcntMaskSlice STEXT size=165 args=0x30 locals=0x18 funcid=0x0 align=0x0
github.com/bits-and-blooms/bitset.popcntAndSlice STEXT size=158 args=0x30 locals=0x18 funcid=0x0 align=0x0
github.com/bits-and-blooms/bitset.popcntOrSlice STEXT size=158 args=0x30 locals=0x18 funcid=0x0 align=0x0
github.com/bits-and-blooms/bitset.popcntXorSlice STEXT size=158 args=0x30 locals=0x18 funcid=0x0 align=0x0

NEW:
can inline popcntMaskSlice with cost 26
can inline popcntAndSlice with cost 26
can inline popcntOrSlice with cost 26
can inline popcntXorSlice with cost 26

github.com/bits-and-blooms/bitset.popcntMaskSlice STEXT nosplit size=82 args=0x30 locals=0x18 funcid=0x0 align=0x0
github.com/bits-and-blooms/bitset.popcntAndSlice STEXT nosplit size=79 args=0x30 locals=0x18 funcid=0x0 align=0x0
github.com/bits-and-blooms/bitset.popcntOrSlice STEXT nosplit size=79 args=0x30 locals=0x18 funcid=0x0 align=0x0
github.com/bits-and-blooms/bitset.popcntXorSlice STEXT nosplit size=79 args=0x30 locals=0x18 funcid=0x0 align=0x0