Better inlining

[lemire]

At the cost of having a larger roaring.h file, this PR significantly improves the performance of roaring_bitmap_contains.

Before

-----------------------------------------------------------------------
Benchmark             Time             CPU   Iterations UserCounters...
-----------------------------------------------------------------------
RandomAccess       5359 ns         5351 ns       128009 GHz=3.25364 cycles=10.955k instructions=33.398k

After

-----------------------------------------------------------------------
Benchmark             Time             CPU   Iterations UserCounters...
-----------------------------------------------------------------------
RandomAccess       2758 ns         2754 ns       203315 GHz=3.2813 cycles=7.652k instructions=22.296k

By allowing more inlining of container_iterator_next and container_iterator_prev, it also helps the performance of other functions. I get that IterateAll64 is faster now.

Before

-------------------------------------------------------
Benchmark             Time             CPU   Iterations
-------------------------------------------------------
IterateAll64    3045904 ns      3034746 ns          244

After

-------------------------------------------------------
Benchmark             Time             CPU   Iterations
-------------------------------------------------------
IterateAll64    2700777 ns      2698000 ns          277

It also seems to help the RandomAccess64Cpp benchmark, but not that RandomAccess64 results. So roaring64_bitmap_contains is NOT improved much (if at all) by this PR.

Before:

------------------------------------------------------------
Benchmark                  Time             CPU   Iterations
------------------------------------------------------------
RandomAccess64          3086 ns         3083 ns       228588
RandomAccess64Cpp       2110 ns         2109 ns       329877

After:

------------------------------------------------------------
Benchmark                  Time             CPU   Iterations
------------------------------------------------------------
RandomAccess64          2970 ns         2967 ns       221617
RandomAccess64Cpp       1801 ns         1799 ns       409146

Fixes: #755

[lemire]

cc @madscientist

[SLieve]

Looks good, thanks!

[lemire]

@Dr-Emann ? I'd like to have some kind of community consensus before merging this as it is quite a change.

[lemire]

@Dr-Emann Thanks for the excellent review.