338 request for fuselets

[anyzelman]

MR that provides so-called fuselets.

Fuselets are sequences of operations that may occur in existing code bases, and could be accelerated using the nonblocking backend. A fuselet represents such a sequence of operations, which are exposed as a single plain-C function call. ALP during its standard build will generate a library of those fuselets using its nonblocking backend, thus resulting in high-performance, automatically fused kernels.

The fuselets furthermore, at run-time and again automatically, tunes performance parameters such as tile sizes and the number of threads to be used.

Finally, the example fuselets provided in include/transition/fuselets.h and src/transition/fuselets.cpp demonstrate how easy it is to use ALP to provide arbitrary fuselets that existing code bases may require, and are easily modified or extended.

This MR also includes:

• smoke and performance tests for the new fuselets;
• standard semiring and monoid definitions in the new grb::monoids and grb::semirings namespaces;
• tests for those standard semirings and monoids (insofar possible);
• a bugfix in the stdout of the spmspm unit test;
• a bugfix for missing cstdint includes in various headers; and
• as always, minor code style fixes.

[anyzelman]

Code review nearly complete. Could remove some element-wise scalar divisions, but need to double-check (lack of) impact on performance. After that, should be ready for merge.

[anyzelman]

Code review complete. Waiting CI OKs before merge.

[anyzelman]

All CIs pass, manual full test suite runs OK, manual checks of fuselets speedup (vs. blocking) also pass. Will merge.