Testing 'n' fixes

[mumbleskates]

Mostly some spit & polish changes:

• run the build ci step in the master branch
• run cargo test --benches for sanity in the build step, with dev mode debug assertions
  • this catches problems like some unaligned pointer reads & an arithmetic panic in the rng implementation of one of the test data types
• general organization in Cargo.toml, some alphabetization and grouping features together a bit to make things simpler
• i also propose here to disable wiring from the default bench set after some discoveries i made chasing down what was going on with the aforementioned unaligned pointer read (which is to my knowledge already UB since it's done through a reference); in trying to whip up a test case for that i found a logic bug that could be an even more serious problem.

[djkoloski]

louaykamel/wiring#3 is the wiring issue for posterity