Support the Omega time coordinate

[xylar]

This merge fixes the manufactured solution analysis task in Omega by using the time coordinate instead of the MPAS-Ocean xtime variable.

It also attempts to use time in the viz step but testing is currently inhibited by #248

Checklist

• Testing comment in the PR documents testing used to verify the changes

[xylar]

Needs #231 and E3SM-Project/Omega#169

[xylar]

I was able to run the manufactured solution test case successfully, including analysis. However, without E3SM-Project/Omega#170, it doesn't behave correctly (as expected) and the convergence rate is terrible.

[xylar]

Viz should be tested after #250 is merged and this gets rebased.

[xylar]

Testing

Tested with manufactured solution on Chrysalis, see #231 (comment)

[cbegeman]

Suggested change

	# Omega v0 uses stacked shallow water where ssh as nVertLevels
	# Omega v0 uses stacked shallow water where ssh has nVertLevels

[cbegeman]

Suggested change

	if model == 'mpas-o':
	tidx = time_index_from_xtime(ds_out.xtime.values, time)
	else:
	# time is seconds since the start of the simulation in Omega
	tidx = np.argmin(np.abs(ds_out.Time.values - time))
	if model == 'mpas-o':
	dt = time_since_start(ds_out.xtime.values)
	else:
	# time is seconds since the start of the simulation in Omega
	dt = ds_out.Time.values
	tidx = np.argmin(np.abs(dt - time))

Can we break up time_index_from_xtime into two functions:

def time_index_from_xtime(xtime, dt_target, start_xtime=None):
    dt = time_since_start(ds_out.xtime.values, start_xtime=None)
    time_index = np.argmin(np.abs(np.subtract(dt, dt_target)))
    return time_index

def time_since_start(xtime, start_xtime=None):
    if start_xtime is None:
        start_xtime = xtime[0].decode()

    t0 = datetime.datetime.strptime(start_xtime, '%Y-%m-%d_%H:%M:%S')
    dt = np.zeros((len(xtime),))
    for idx, xt in enumerate(xtime):
        t = datetime.datetime.strptime(xt.decode(),
                                       '%Y-%m-%d_%H:%M:%S')
        dt[idx] = (t - t0).total_seconds()
        return dt

[cbegeman]

I suggested a minor change, which you can take or leave. I successfully ran the analysis step of manufactured_solution with this change.