Improving MODIS tile handling

[sigmafelix]

• When MODIS tiles are downloaded by tile indices, there is a case where only a few of the tiles are available in the index range.
• An example is 2021.09.22 MOD09GA products, when the mainland US is completely missing even though the upper left corner (h08v03) in the Northern Pacific Ocean remained due to the range setting (h: 8-13, v: 3-6). It returned an error because of the collision between NULL return and the postprocessing functions for data.frame, leading to failing the calculation at other dates.
• Better logic should be implemented; a quick idea is --
  • [ ] to use standard deviation or median of the number of tiles then to exclude dates when the number is below a preset threshold or aforementioned statistics.

[sigmafelix]

I decided to use mode to get the typical number of files from other file path strings. Any dates with less than 80% of the mode of the number of files in other days will be excluded. It still has a limitation that file paths are used as they are when the number of days are too low. However, since we assume calc_modis_par() is supposed to get many days worth data paths in from argument to take advantage of parallel processing, the impact of this limitation is trivial.