Add check for negative size when generating intervals

Minor fixes for local smoothing

Add make_cal.py script to generate CAL format from observation web se…

Better default l0 in additive() method, etc.

fixed indentation

Previously, if l0 was None, we tried to calculate a default initial level. This was
problematic if yobs had no valid points from which to calculate a mean. Now, if yobs
has no valid points, default to l0=0. This is a bit arbitrary, but it is necessary
to prevent a NaN-valued l from being passed back out. Other defaults are equally
arbitrary.

Also, I noticed that the `s` variable was getting converted from a NumPy array to a
list just before the method returned. For consistency with other outputs, we no
longer convert, and just return a NumPy array.

Pad to urlInterval in put_timeseries()

I'm not sure this is functionally different from what there was previously. Is this what you were suggesting in your comment on this PR @jmfee?

TimeseriesFactory.get_timeseries() pads to the requested interval with NaNs,
filling in start/end gaps. TimeseriesFactory.put_timeseries() is supposed to
do this (according to Jeremy Fee), but it was trimming NaNs from the start/
end before writing to file. This PR fixes this by using ObsPy's Stream.trim()
method, in a similar manner to TimesereisFactory.get_timeseries(), to pad to
the specified start and end of the urlInterval..

Issue#134

Patch waveserver

Update to MetaData.py based on recent measurements

We recently reviewed our various data products and found discrepancy in some of the metadata.  We have updated these values based on recent measurements.  I've modified these values in ObservatoryMetadata.py to be consistent with our other data products.

- updated init and process_one functions to handle "smooth" parameter
- included a *very* minor fix to updates in Sigma in additive function

SqDist local smoothing

By default, each element of the SQ vector (i.e., shat returned from additive())
is treated as if it is independent of its nearest neighbors in time. This is
not generally true, but to treat this in a completely statistically robust
manner would be computationally prohibitive. Instead, we apply a Gaussian
smoother/filter to a limited set of SQ elements in the neighborhood of the
current time step. The configuration parameter 'smooth' specifies the period
(in samples) at which one requires the amplitude to be attenuated by half.
Longer periods will be less attenuated, while shorter periods will be more
attenuated.

Also, added the running estimate of the standard error (i.e., sigmahat) to
the stream returned by process_one().

Added command-line option to set Z-score threshold

Fixes #131