Add downsapmling for ceresdb

[chenmudu]

Describe This Problem

At this stage, downsampling has become one of the basic capabilities in the field of time-series databases, which can solve problems such as excessive data volume returns under certain special circumstances (large time interval, high data set density).

CeresDB should support such a capability, at least in the immediate plans.

Proposal

The engine layer supports downsampling, and supports dynamic downsampling and configuration downsampling. Matching, clients in each language should also support such an API.

Notice：

• dynamic downsampling：Support can automatically adjust the sampling granularity according to the time interval.
• configuration downsampling：Rely on the user to provide the accuracy of the downsampling.

Additional Context

OpenTSDB Downsampling

[jiacai2050]

Thanks for suggestion.

Downsample is indeed important for time-series user-case, to support downsample an RFC maybe required to discuss

• pros and cons of downsample
• when query will use downsampled data
• how to run downsample job(in main server or some other background components).
• ....

[chenmudu]

• pros and cons of downsample

In most people's opinion, downsampling should have no disadvantages for time series data. Maybe I'm missing something, can anyone add here.