[BUG]: Y range of the loss chart in the Calibrate operator is too wide - should be relative to the data, not start from zero always

[liunelson]

Describe the bug

In the Calibrate operator output panel, there is a "Loss" line chart. The y-axis range (min, max) should be based on the min/max range of the loss dataset, instead of always having the min = 0.

The reason is that the absolute value of the loss has no inherent meaning - the user only needs to know whether the loss as a function of iterations is "well-behaved". It is well-behaved when it decreases monotonically. The current chart with a fixed Y min = 0 prevents this observation when loss >> max(loss) - min(loss).