Make caching smarter by honoring data publication schedule

[amotl]

Is your feature request related to a problem? Please describe.
When working on #243, we should also think about making the caching layer smarter. With multiple invocations of --period=historical, for example, data should not be downloaded over and over again.

Describe the solution you'd like
Wetterdienst should honor DWD publication times in order to save resources.

For implementing that, the program must know exactly about the publication schedules of each data period. Letting Wetterdienst know about DWD publication times and update definitions also resonates with #225 somehow.

[wetterfrosch]

Talking about smartness:
When I query eg the past year from historical (2019-01/2019-12), which is within the 10-year timerange of a single file, all historical files are loaded: Currently I count up to three, beginning in 1992. This doesn't seem efficient. Are all of these values loaded into the RAM?

[gutzbenj]

Dear @wetterfrosch , an update is on its way to manage historical data in this context more smarter.

[gutzbenj]

This is implemented now by #267

• this now uses dynamic periods depending on start date and end date
• high resolution data with multiple files is filtered for only those files which are overlapping with start date and end date