utilities: Update time/r methods for LINUX

Default to MONOTONIC clock for timer methods to avoid falling victim to
clock corrections.  Changed signatures from accepting pointers since
this is not needed an complicates calls and Java/JS/Python bindings.

    * Switched from nanosleep to clock_nanosleep to allow developers to
      provide a clock for LINUX
    * Default upm_clock_init to CLOCK_MONOTONIC
    * Updated logic to calculating delay and elapsed to be more readable
    * Added ns flavors for completeness
    * Refactored all upm_* delay/timer methods
    * Added #else for preprocessor cases w/o an #else
    * Added test for AQI
    * Added test fixture with logic to identify a minimum delay time
      which is used as a metric for testing all delay methods
    * Much more lenient unit testing of delays to minimize false CI
      failures

Signed-off-by: Noel Eck <noel.eck@intel.com>

src/speaker/speaker.c

@@ -206,8 +206,7 @@ upm_result_t speaker_emit(const speaker_context dev, unsigned int freq,
     if (speaker_set_frequency(dev, freq))
         return UPM_ERROR_OPERATION_FAILED;
 
-    upm_clock_t clock;
-    upm_clock_init(&clock);
+    upm_clock_t clock = upm_clock_init();
 
     mraa_pwm_enable(dev->pwm, 1);
     while (upm_elapsed_ms(&clock) < emit_ms)