optimized range insertion for boost::container::hub

[joaquintides]

Optimizes range insertion taking advantage of the fact that insertion goes through all non-empty and non-full blocks first and then uses empty blocks exclusively. Current hub tests cover all the new code, including catch blocks.

Performance increase (perf.cpp, release mode):

GCC 13

Before

n       t(regular insert) / t(range insert)
10E2    1.1193
10E3    1.14559
10E4    1.14469
10E5    1.18275
10E6    1.16951

After

n       t(regular insert) / t(range insert)
10E2    1.67431
10E3    1.38766
10E4    1.42388
10E5    1.71183
10E6    1.7539

Clang 13 for Visual Studio (clang-cl)

Before

n       t(regular insert) / t(range insert)
10E2    1.16093
10E3    1.27721
10E4    1.1313
10E5    1.18809
10E6    1.2059

After

n       t(regular insert) / t(range insert)
10E2    1.65984
10E3    1.916
10E4    1.4827
10E5    1.6352
10E6    1.70056

VS 2022

Before

n       t(regular insert) / t(range insert)
10E2    1.25101
10E3    1.26594
10E4    1.16627
10E5    1.09216
10E6    1.21067

After

n       t(regular insert) / t(range insert)
10E2    1.41975
10E3    1.44301
10E4    1.30676
10E5    1.32164
10E6    1.38581