Segyio

* segyio has learned to understand rev2 1-byte header words
* segyio now understand rev2 wide header word replacements
* segyio now detects traces with >65k samples
* Support for int24 formats
* Fixed a bug where f.trace[i,j] returned an array even when i,j are ints
* SEGY_INVALID_SORTING message has been improved
* SEGY_SIGNED_INTEGER_3_BYTE enum alias for symmetry with signed counterpart
* Minor build system and CI improvements

* Text headers with null-bytes are no longer truncated
* Slowness on sparse writes is pointed out in documentation
* The python package now respects CMAKE_INSTALL_PREFIX
* Multibuild is now cloned on travis, not a submodule
* tools.wrap now works in python3, by decoding the string
* A few C compiler warnings addressed
* collections.abc is imported a future-proof way in gather.py

* segyio has learned to work with with >32k samples per trace
* collections.abc is imported a future-proof way
* Support for python 3.4 has been dropped, as it is end-of-life
* Python 3.8 support

* Build and packaging fixes

* A new feature, group, has been added with experimental support (hence no
minor release). This is intended for pre-stack files, and is useful for
grouping up traces by shot gathers and similar
* Segyio has learned how to read subtraces, f.trace[trno, first:last]
* Source water depth is now correctly a 4-byte word
* Fixed a bug with header-assignments in pre-stack files
* A gallery has been added to the docs, for rich examples
* Various documentation improvements, examples, and spelling corrections
* The python build system is rewritten to use scikit-build
* The exported cmake targets are namespaced with segyio::
* segyio now requires CMake >= 3.5 (from 2.8.12) to build

* Fixed a bug where negative short header words on gcc were read wrong