Mixing temporal resolutions when merging files with combine_first

[BaptisteVandecrux]

If an hourly STM table covering 2022-2024 is retrieved from the field and is added after a raw 10 min file covering 2022-2023, then the merging of L1 datasets into the L1A dataset will have a strange behavior:

pypromice/src/pypromice/process/aws.py

Line 138 in 3813c77

self.L1A = reduce(xr.Dataset.combine_first, reversed(self.L1))

The latest hourly data ds_latest_hourly = AWS.L1[-1] will be merged with the previous raw dataset ds_previous_10min = AWS.L1[-2] with :
ds_merged = ds_latest_hourly.combine_first(ds_previous_10min )

In that statement, the round-hours lines from the hourly file will be unchanged, but 10 min data will be inserted between these values.

[BaptisteVandecrux]

Due to this this, winter daily tx have been injected at KAN_U's hourly data:

Again, it all comes from the fact that combine_first only looks at the order at which the files come and is not setting different priorities to the different data types (10 min > STM > tx).