* rename get_size to axes_rank and move axes buffer to axes.hpp
* added fill experiments
* faster and simpler fill_impl
* faster specializations for linearize_value
- Tests and benchmarks now only include the headers they really need.
- Warnings turned off for tests that compile with external boost libs that are not warning free
* InputIterator concept requires that the decrement of a postfix-incremented iterator is valid, now it is
* added detached_accessor to that end
* update benchmark plot, better docs
* fetching google benchmark with cmake
* use new thread-safe accumulator in histogram_parallel_filling
* update boost, root and gsl benchmarks to use google benchmark
* fixed allocator-failing code with better mock allocator; 100 % code coverage on all platforms
* fixed and refined use of safe comparison of number types
* provide unsafe access to buffer
* factored mp_int out into separate header and renamed to large_int
* refined adding behavior
* more warnings and comeback of coverage variant
* better coverage reporting
- new thread-safe accumulators for arithmetic types
- storages have a new flag `has_threading_support` to indicate support for parallel cell access
- histogram automatically synchronises growing axes and storage if necessary
* concept docs rewritten, adapting to tabular style as in Beast and cppreference.com
* update the readme with correct links and text
* removed necessity to add `storage_tag` member type to storage classes
* better operators for mp_int, unlimited_storage::reference, map_impl::reference, copying code from boost.operators
* treat warnings as errors on travis
* make appveyor build feature branches
* fixes broken msvc build
* smaller xml files for testing serialization
* fixing many warnings on msvc
* bug-fixes to mp_int regarding operators and new tests to check their behavior
* don't expect iterable histogram::at to have size method
* build benchmarks in release mode
* test everything by default when b2 is run on project directory
* more test coverage
* Travis uses b2 and codecov now
* replacing boost::multiprecision with custom implementation to avoid the dependency
* improved axis implementation: `update` is now a normal method
* introduced use_default tag type to set defaults
* whitespace fixes and doc update
* renaming naked to remove_cvref_t
* more static_asserts to check implicit conditions
* use new BOOST_TEST_TRAIT_SAME
* reorganize and add operator tests, fixes for array and map adaptors
* improved map adaptor
* storage does not have to support scaling operator anymore
* doc improvements
