Better date handling with missing LS8 bands

[dshean]

Fixes error that I encountered using LS8 row,path (46, 27) with example notebook. There was one missing B4.TIF file, which produced inconsistent number of datasets and dates (when extracted from original df).

[dshean]

Now realizing this might not be repo for latest example notebooks. @scottyhq, maybe you can integrate wherever appropriate?

[scottyhq]

Sorry for the confusion @dshean, there are now a few versions of this notebook with a few fixes as you've picked up. the search was set up without date bounds and it seems like that was causing the issue. See also a slightly modified notebook from the December 2018 AGU tutorial (https://github.com/pangeo-data/pangeo-tutorial-agu-2018), and a demonstration of accessing the same Landsat8 data on AWS as of Feb 2019 (https://github.com/scottyhq/esip-tech-dive).

[dshean]

OK, thanks for the links. It's still an issue in the updated notebooks. Issue is:
index = gpd.pd.DatetimeIndex(items.dates(), name='time')
You're assuming all items have valid datasets. My case with a different path/row had one missing B4.tif, so index contained 112 items.dates(), but datasets lists contained only 111 datasets from the for loop in the cell above with the try/except. So the mismatch causes concatenate step to fail (da = xr.concat(datasets, dim=index)). I just created another list to store dates and used that to create the Datetimeindex, but could also pull out directly from the datasets list. Should be a simple fix.