Add filtering options for layers

[nyalldawson]

Add an equivalent of a vector layer's "subset string" type filtering, where we can restrict the points rendered in both 2d and 3d views by elevation range and classification

Basically we would ensure that the point cloud blocks returned by the provider implement this filtering internally, so that the returned blocks only contain the desired attributes/z range points

[wonder-sk]

Shall we reuse QgsExpression for that?

[nyalldawson]

I was initially thinking of skipping any textual representation of a filter for now and just adding the underlying class structure representing it (and constructimg that direct from the gui). I'm not sure how well QgsExpression would fit here... Let me think on that one!

[wonder-sk]

Just for inspiration / reference...

Filtering in PDAL - using a dedicated filter - https://pdal.io/stages/filters.range.html#filters-range

Filtering in PDAL - in the latest version most (or all?) filters support where clause for filtering of points based on expressions: https://pdal.io/stages/expression.html

Filtering in LASTools - command line arguments:

Filter points based on their coordinates.
  -keep_tile 631000 4834000 1000 (ll_x ll_y size)
  -keep_circle 630250.00 4834750.00 100 (x y radius)
  -keep_xy 630000 4834000 631000 4836000 (min_x min_y max_x max_y)
  -drop_xy 630000 4834000 631000 4836000 (min_x min_y max_x max_y)
  -keep_x 631500.50 631501.00 (min_x max_x)
  -drop_x 631500.50 631501.00 (min_x max_x)
  -drop_x_below 630000.50 (min_x)
  -drop_x_above 630500.50 (max_x)
  -keep_y 4834500.25 4834550.25 (min_y max_y)
  -drop_y 4834500.25 4834550.25 (min_y max_y)
  -drop_y_below 4834500.25 (min_y)
  -drop_y_above 4836000.75 (max_y)
  -keep_z 11.125 130.725 (min_z max_z)
  -drop_z 11.125 130.725 (min_z max_z)
  -drop_z_below 11.125 (min_z)
  -drop_z_above 130.725 (max_z)
  -keep_xyz 620000 4830000 100 621000 4831000 200 (min_x min_y min_z max_x max_y max_z)
  -drop_xyz 620000 4830000 100 621000 4831000 200 (min_x min_y min_z max_x max_y max_z)
Filter points based on their return numbering.
  -keep_first -first_only -drop_first
  -keep_last -last_only -drop_last
  -keep_second_last -drop_second_last
  -keep_first_of_many -keep_last_of_many
  -drop_first_of_many -drop_last_of_many
  -keep_middle -drop_middle
  -keep_return 1 2 3
  -drop_return 3 4
  -keep_single -drop_single
  -keep_double -drop_double
  -keep_triple -drop_triple
  -keep_quadruple -drop_quadruple
  -keep_number_of_returns 5
  -drop_number_of_returns 0
Filter points based on the scanline flags.
  -drop_scan_direction 0
  -keep_scan_direction_change
  -keep_edge_of_flight_line
Filter points based on their intensity.
  -keep_intensity 20 380
  -drop_intensity_below 20
  -drop_intensity_above 380
  -drop_intensity_between 4000 5000
Filter points based on classifications or flags.
  -keep_class 1 3 7
  -drop_class 4 2
  -keep_extended_class 43
  -drop_extended_class 129 135
  -drop_synthetic -keep_synthetic
  -drop_keypoint -keep_keypoint
  -drop_withheld -keep_withheld
  -drop_overlap -keep_overlap
Filter points based on their user data.
  -keep_user_data 1
  -drop_user_data 255
  -keep_user_data_below 50
  -keep_user_data_above 150
  -keep_user_data_between 10 20
  -drop_user_data_below 1
  -drop_user_data_above 100
  -drop_user_data_between 10 40
Filter points based on their point source ID.
  -keep_point_source 3
  -keep_point_source_between 2 6
  -drop_point_source 27
  -drop_point_source_below 6
  -drop_point_source_above 15
  -drop_point_source_between 17 21
Filter points based on their scan angle.
  -keep_scan_angle -15 15
  -drop_abs_scan_angle_above 15
  -drop_abs_scan_angle_below 1
  -drop_scan_angle_below -15
  -drop_scan_angle_above 15
  -drop_scan_angle_between -25 -23
Filter points based on their gps time.
  -keep_gps_time 11.125 130.725
  -drop_gps_time_below 11.125
  -drop_gps_time_above 130.725
  -drop_gps_time_between 22.0 48.0
Filter points based on their RGB/CIR/NIR channels.
  -keep_RGB_red 1 1
  -keep_RGB_green 30 100
  -keep_RGB_blue 0 0
  -keep_RGB_nir 64 127
  -keep_NDVI 0.2 0.7 -keep_NDVI_from_CIR -0.1 0.5
  -keep_NDVI_intensity_is_NIR 0.4 0.8 -keep_NDVI_green_is_NIR -0.2 0.2
Filter points based on their wavepacket.
  -keep_wavepacket 0
  -drop_wavepacket 3
Filter points based on extra attributes.
  -keep_attribute_above 0 5.0
  -drop_attribute_below 1 1.5
Filter points with simple thinning.
  -keep_every_nth 2 -drop_every_nth 3
  -keep_random_fraction 0.1
  -thin_with_grid 1.0
  -thin_pulses_with_time 0.0001
  -thin_points_with_time 0.000001
Boolean combination of filters.
  -filter_and

LASTools GUI for filtering:

[NEDJIMAbelgacem]

This relates to https://github.com/qgis/QGIS/pull/40464 a bit Now I'm doing the filtering in the shader by discarding the pixel that is not selected among the categories. The discard instruction causes a lot of warp divergence which results in a big performance hit.