Add shot_id as a dimension to datasets for combining data.

[samueljackson92]

A great suggestion by Greg from PSFC: we should add "zero size" dimension to each dataset called shot_id who's value is the shot id.
This would allow the user to combine different datasets by concatenating along the shot_id axis.

[LariDG]

When adding the shot_id as a dimension found that the time bases are not aligned between shots. Time is interpolated based on time-step and as the shots have different start times and time ranges the time series don't match up exactly. Hence, when concatenating via the new shot_id many nans are formed.

For example:

{30420, summary}
Dimensions: (shot_id: 1, time: 1743)
{30420, summary}.time
array([-0.0568 , -0.05655, -0.0563 , ..., 0.3782 , 0.37845, 0.3787 ])

{30421, summary}
Dimensions: (shot_id: 1, time: 2906)
{30421, summary}.time
array([-0.0612 , -0.06095, -0.0607 , ..., 0.66455, 0.6648 , 0.66505])

Then after comb = xr.concat([ds_20_sum, ds_21_sum], dim="shot_id")

Dimensions: (shot_id: 2, time: 4649) (has 1743 + 2905 time points)
Showing that the time coordinates do not overlap, hence, there will nans filled in for the values of the respective time coordinates that there is no data for:

shot_id   30420        30421
time                        
-0.06120    NaN        -6039.37
-0.06095    NaN        -6346.67
-0.06070    NaN        -6516.91
-0.06045    NaN        -6626.93
...
-0.05695    NaN        -5638.63
-0.0568     -5458.02   NaN
-0.0567     NaN        -5509.46
-0.05655    -5175.29   NaN
-0.05645    NaN        -5381.70
....
0.0665      NaN        427.54

Time bases are similar, so when plotting this isn't noticeable:

[samueljackson92]

I think we should make a new issue and change this to be consistent. Code is here. Broadly, for L2 I think we should pick and fix min value (e.g. -0.1), and then interpolate to the max time (rounded up to the nearest bin width).