* Move headers to relevant include directories within the target tree(s)
* Adjust doxygen configuration to search correct paths
* Adjust CMake configuration to use new include paths

* Replaces conda_* and pip_* with pkg_index_provides
* Because this function can fail in so many ways I've added pkg_index_provides_strerror() and a detection macro PKG_INDEX_PROVIDES_FAILED() to make things easier

* To be more aligned with conda_provides, the version spec has been consolidated into one argument

* Add python_package_exists() function

* Poll pypi.org or compatible index to see if a package exists
* Returns non-zero on success

* Implements python_package_exists() in delivery_defer_packages()

* Implements python_package_exists() in delivery_defer_packages()

* Bugfix: Avoid incorrect package selection

* With large package lists that contain multiple packages starting with the same strstr() would pick the first match
* This adds a temporary name variable that strcmp() can check against.

* Message correction:

* Change "release" to "testing" in testing environment failure message

* Amend message to fit the flow of the output

* Disable outdated conda notifications

* The latest version isn't always the greatest. Don't give the end-user any ideas. Just use whatever the installer provides... quietly

* Rename python_package_exists to pip_index_provides

* Document the function prototype
* Add missing comments in micromamba structure

* Ensure the temporary output file does not linger

* Change return value of conda_setup_headless() from void to int

* Replace exit() with return;

* Return early if unpacking the micromamba binary fails

* Exit program when pointer to INIFILE is NULL.

* Validation function cannot otherwise proceed

* The way the logic is set up I've decided to duplicate the installation code for now until I find time to revise it

* The only meaningful difference between a "fresh start" and reusing the conda installation is a rmtree().

* Exposes STASIS_DOWNLOAD_TIMEOUT environment variable

* Sets the connection timeout for libcurl to 30, instead of 300.

* Export ini_section_create() function

* Add download() tests

* Add conda_*() tests

* Add boilerplate source file for test framework

* Fixes segfault reported by @GeorgeJCleary (#10)

* The key is now an array index. When key is -1, the env variable is not defined.

* Free resources only when continue on error is disabled (#11)

* Fix segfault due to premature shutdown/cleanup

* If conda_setup_headless cannot succeed, die

* Set STASIS_SYSCONFDIR for tests

* Pass .ci_support/plat_arch_.yaml to conda-build

* Fixes a few outstanding leaks in delivery context

* Move micromamba function out of stasis_indexer.c

* Adjust code in the indexer to accommodate the move. The function now expects a MicromambaInfo structure as its first argument.

* Add missing warning message

* User is informed when pandoc is not available for HTML page generation

* Initialize workdir_template string to zero

* Add micromamba program to runtime PATH

* Expose storage.tools_dir to template engine

* Remove dead code

* Fix wording in comment

* Fix conda-forge builds

* Pass their .ci_support configurations to conda-build in order to fully set up their build runtime environment

* Add get_cpu_count()

* Exposes STASIS_CPU_COUNT and CPU_COUNT to the runtime environment

* Implements conda reactivation template string

* {{ workaround.conda_reactivate }}
* This is useful to call after installing any conda packages within a test.script

* Fix conda runtime inside of test.script

* This ensures conda and mamba are fully initialized.
* Previous behavior only placed the commands on the PATH but didn't provide any shell macros (i.e. undefined behavior)

* Document CPU_COUNT and workaround.conda_reactivate