Abstract framework for tree navigation.  Based on this framework, specific libraries for
particular hierarchical data can be implemented, for example, file systems.

A typical tree would look like the following:

  [+] Node 1
  [-] Node 2
   |    [+] Node 21
   |    [+] Node 22
   |    [-] Node 23
   |     |    [+] Node 231
   |     |      Leaf 231
   |     |      Leaf 232
   |     |      Leaf 233
   |    [+] Node 24
  [+] Node 3
  [+] Node 4

Each node has a "label".  Non-leaf nodes also have a "control".  The label is the text by which
the node is represented.  The control is a clickable piece of text in front of the label which
allows the user to fold or unfold the node.  Nodes without children usually don't have a control.
Nodes can also have an icon, but this is optional.  Icons are implemented as images or symbols in
an icon font.  Control, icon, and label are always displayed in this order, and are always
displayed in one line.  Thus, multi-line nodes are not supported by this library.

Nodes are represented by lists of the form:

  (NAME PROPS PARENT CHILDREN)

where the elements have the following meaning:

  NAME      The name of the node.  A string.

  PROPS     The properties of the node.  A plist.  See below for details.

  PARENT    The parent of the node.  Another node-representing list.

  CHILDREN  The children of the node.  List of node-representing lists.

Node properties are implemented as plists with Lisp symbols as keys.  The following
properties exist:

  node-line-overlay
         The overlay containing the node line

  indent-overlay
         The overlay containing the node indentation

  label-overlay
         The overlay containing the node label

  control-overlay
         The overlay containing the node control

  icon-overlay
         The overlay containing the node icon

  start  A marker at the position where the node begins

  end    A marker at the position where the node ends

  state  The state of the node.  See below for details

  selected
         Whether the node is selected (non-nil if selected, nil if not)

Node states: Each node is in exactly one of three states, which are represented by the
following Lisp symbols:

  folded-unread   The node is folded and has not been unfolded before

  folded-read     The node is folded, but has been unfolded before

  expanded        The node is expanded

Initially, a node is in the state 'folded-unread'.  If it gets expanded for the first time,
the state changes to 'expanded'.  If it gets folded again, the state changes to 'folded-read'.
From then, the state toggles between 'expanded' and 'folded-read' each time the node is
expanded or folded.

The framework is used by implementing several functions which are defined as function variables,
thus, variables whose values are function symbols.  Here is a list of that variables:

  treeview-get-root-node-function
  treeview-node-leaf-p-function
  treeview-update-node-children-function
  treeview-after-node-expanded-function
  treeview-after-node-folded-function
  treeview-get-indent-function
  treeview-get-control-function
  treeview-get-control-margin-left-function
  treeview-get-control-margin-right-function
  treeview-get-icon-function
  treeview-get-icon-margin-left-function
  treeview-get-icon-margin-right-function
  treeview-get-label-function
  treeview-get-label-margin-left-function
  treeview-get-control-keymap-function
  treeview-get-indent-face-function
  treeview-get-control-face-function
  treeview-get-control-mouse-face-function
  treeview-get-selected-node-face-function
  treeview-get-highlighted-node-face-function
  treeview-get-label-keymap-function
  treeview-get-label-face-function
  treeview-get-label-mouse-face-function
  treeview-get-icon-keymap-function
  treeview-get-icon-face-function
  treeview-get-icon-mouse-face-function
  treeview-suggest-point-pos-in-control-function

A description of each variable can be found in the repsetive documentation strings.  All
variables are buffer-local.  Libraries using this framework should create a new buffer and
set the variables to particular functions in that buffer.  Then, the root node should be
created and rendered in the buffer by a call to treeview-display-node.

See library "dir-treeview" for an example where this framework is used.

If you upgrade from v1.0.0: Note that since v1.1.0 another buffer-local function variable
treeview-get-root-node-function exists which didn't exist in v1.0.0. It must be set to a
function which returns the root node of the tree.