google#639)

…ndering with unequal chunk sizes (google#632)

* fix: pass correct chunk size to 3D histogram

* feat: more accurately distribute samples in 3D into chunks for histogram calculation

* fix: remove discard in 3D histogram for display purposes
this means the histogram now reflects the underlying data more

* refactor: cleaner var names and sum

* refactor: clearer variable naming for histogram instances

* fix: formatting

…rlHash (google#644)

without a maxWait value, debounce can cause important state changes to be lost if there is high rate of state change between the important change and a page reload
mouse movement causes state changes and is probably the main reason this can happen

TensorStore encodes chunks of shape (x_size, y_size, z_size, c_size)
with width=x_size*y_size, height=z_size, while cloud-volume uses
width=x_size, height=y_size*z_size.

With this change, Neuroglancer will accept any width/height as long as
`width * height = x_size * y_size * s_zize`.  This loose checking was
already used for the JPEG format.

Fixes google/tensorstore#196

* fix: correct deletion of old queries

* feat(framerate): framerate fixes when application errors occur

* feat: store the frame number with the delta

* refactor: easier management of the query pool

…er (google#645)

…en hovering inside the ikonate svg with a default title (google#633)

…ogle#656)