Add period values for 'sec' and 'ms' granularities, as it appears the…

…y are missing.

If the granularity detection results in "sec" or "ms" granularity
detect_anons will blow up with a message stating 'period' was not
provided. Which is correct. This provides at least some defaults.

This is a bit tricky, since the number of samples really isn't guaranteed
to to be 1 measurement per whatever the 'gran' variable says it is.
However, this code appears to assume that. Leading to "fun" when the
precision is in ms but the measurements fire a lot less often - say, once
per five minutes (that's the testcase in question).

There are two ways to deal with this, neither of them implemented by this patch:
- Look at the delta between the first and the last timestamp and the number
of records and make an educated guess from there.
- Add 'period' as a function argument and let the caller decide, using the
hardcoded values simply as educated guesses.

These approaches can be combined, but that is left as an exercise for the reader.

R/ts_anom_detection.R

@@ -166,11 +166,15 @@ AnomalyDetectionTs <- function(x, max_anoms = 0.10, direction = 'pos',
     x <- format_timestamp(aggregate(x[2], format(x[1], "%Y-%m-%d %H:%M:00"), eval(parse(text="sum"))))
   }
 
+  ## This is a bit tricky, since the number of samples really isn't guaranteed to be 1 measurement per whatever the 'gran' variable says it is.
+  ## Either we'll need to do something smarter (look at the delta between the first and the last timestamp and count the number of rows) or
+  ## alternatively, simply make 'period' a function argument and use these values as defaults.
   period = switch(gran,
                   min = 1440,
+                  ms = 1000,
+                  sec = 60*60,
                   hr = 24,
-                  # if the data is daily, then we need to bump the period to weekly to get multiple examples
-                  day = 7)
+                  day = 7) # if the data is daily, then we need to bump the period to weekly to get multiple examples
   num_obs <- length(x[[2]])
 
   if(max_anoms < 1/num_obs){