Add power transform as an option for scaling the Y–axis

[ant-222]

Currently, Superset suppports only two modes for the scaling of the Y–axis: linear (no scaling) and logarithmic. This is a request to support the power function or the more general power transform as a third option. Ideally, Superset could also accept an arbitrary transform that the user would specify in the as an SQL expression.

We have a use case for this in out Kanban control chart. The linear scale is totally inadequate:

The logarithmic scale is better, but still inadequte because the few very short tasks eat up a lot of vertical screen estate, causing the bulk of data and the quantile graph to be squeezed within a rather narrow horizontal limits:

Whereas a simple power funtion y^0.5 results in a more balanced, viewable diagram:

but with the Y-axis incorrectly graduated, because we transform the value rather than the axis scale.

Power transformation (aka Box Cox transformation) is a generalisation of log and power transoforms. Will you please add it as an option for scaing the Y–axis. The user should be able to specify the transormation parameter by hand or have Superset estimate it statistically (e.g. as described here).