[feature request] Store only changes

I think it would be a nice improvement to support the storage of "only changes" in a time-serie.
The "store only changes" is a common downsampling strategy in some environments (e.g. satellite telemetries) and would be nice to have that mechanism native (e.g. without implement continuous query or any other ad-hoc mechanism).

This approach would reduce significantly the number of samples stored in the archive (without missing any information, because some parameters are reported periodically... with the same value once and again)
A continuous flow of insertions like these: 1,2,3,3,4,1,3,3,3,3,3,0,1,1,2 would result in: 1,2,3,4,1,3,0,1,2

The issue names no files or tests. Start by reviewing the existing continuous-query approach and the time-series storage path, then define how repeated values should be detected and persisted. Done means the example sequence is stored as 1,2,3,4,1,3,0,1,2 without losing information.

Written by the indexing model from the issue text.