Dataset2htg

The principle of the filter is to subdivide input data from a vtkDataSet following a subdivision strategy.

This commit implements the algorithm from figure 8 of:

ADR Visualization: A Generalized Framework for Ranking Large-Scale Scientific Data using Analysis-Driven Refinement

authors: Boonthanome Nouanesengsy, Jonathan Woodring, John Patchett, Kary Myers, James Ahrens

The main problem with this algorithm is that it is bottom-up, while hyper tree grids need top-down strategies in order to be able to index each position in the hyper tree grid correctly.

To alleviate this issue, we create a "fake" hyper tree grid without all the indexing properties and the cursors. We build it as a grid of multi-resolution sparse grids stored as std::unordered_map. This is so we do not allocate all the grid. This allows to set rather high MaxDepth parameter for the hyper tree grid while having a small memory footprint.

This grid of multi-resolution sparse grids is filled bottom-up with all the information required, and then we construct the vtkHyperTreeGrid with a top-down recursive algorithm.

Leaves of the hyper tree grid are subdivided following a metric / strategy. A quantity is calculated over all the element of each subtree. In order to avoid calculating the same quantity several times, those measurement are pointer that should inherit from vtkAbstractArrayMeasurement. The measure is carried from the bottom to the top by merging sub vtkAbstractArrayMeasurement objects.

A few measurement methods are implemented in this commit.

Fixes #19213 (closed)