Add performance tests

[jermp]

In a subfolder called perf.

• compact_vector::operator[]
• compact_vector::access
• compact_vector::iterator
• bit_vector::get
• bit_vector::get_bits
• bit_vector::get_word56
• bit_vector::get_word64
• bit_vector::iterator
• darray1::select
• rank9::rank1
• rice_sequence::access
• sdc_sequence::access
• ranked_sequence::access
• endpoints_sequence::locate
• elias_fano::access
• elias_fano::diff
• elias_fano::next_geq
• elias_fano::prev_leq
• elias_fano::locate
• cache_line_elias_fano::access

[jermp]

Note: we should expect endpoints_sequence::locate to be comparable in efficient to endpoints_sequence::access since we store explicitly the positions of the 0 bits and avoid the select. Should be verified emprirically.