* Adding group layer support

* Added test case and fixed parsing code

* Adding comma

* Adding missing newline

* Merging GID changes

* Merging GID changes

* Cleaning files

* Cleaning files

* Cleaning files

* Partial commit

* Partial commit

* Partial commit

* Some suggested changes and fixed tests (except infinite)

* Replaced `tile` member in `LayerTileRef` with a function to get an
  `Option<&Tile>`.

* Replaced `Map::get_tile_by_gid` with `Map::get_tileset_for_gid`, which
  just returns the `Option<&Tileset>`. It also does a reverse search,
  which fixes the lookup in case an external tileset has grown since the
  map was last saved.

* Replaced `Tileset::get_tile_by_gid` with `Tileset::get_tile`, since
  subtracting of the `first_gid` already happens when creating the
  `LayerTileRef`. Also, we eventually should remove `first_gid` from
  `Tileset`, since it should be possible to share a single tileset
  instance betweeen several maps.

* Pre-allocate the tiles hash map for the expected number of tiles and
  use `or_default` instead of `or_insert_with(Default::default)`.

* [nonbuilding] Move ownership of tilesets
- Moves the ownership of tilesets from Map to an object implementing `TilesetCache`

* Clean up

* More cleanup

* Organize layers into modules

* Further modularization

* Add layer wrappers

* Implement `Clone + PartialEq + Debug` for wrappers

* Fix example

* Fix all tests except for test_infinite_tileset

* Move layer utils to its own module

* Better `Map::layers` documentation

* `TilesetCache` -> `cache::ResourceCache`

* Add `ResourcePath`, rename and add errors

* Interface changes
- Move embedded tilesets from cache to map
- Store `Option<LayerTileData>` instead of `LayerTileData`

* parser ->`&mut impl Iterator<Item=XmlEventResult>`

* Document that tilesets are ordered by first GID

* Fix the layer tiles using GIDs issue

* Run `cargo fix`

* Fix `test_infinite_tileset` tests

* Implement a way to access object tile data

* Rename `TiledWrapper` to `MapWrapper`

* More efficient `get_or_try_insert_tileset_with`

* Add `ResourcePathBuf`, use `Rc<Tileset>` in `Map`

* Remove `MapTileset::first_gid`

* Run `cargo fix`

* Remove unrelated `Tileset` changes

* Requested changes

* Avoid reference counting when accessing tiles

* Store tile data instead of (u32, u32)` in objects

* Remove unneeded functions from `ResourceCache`

* Address PR comments

* Misc improvements in `layers::tile`

* Improve example

* Convert `LayerTile` properties to fields

Co-authored-by: Thorbjørn Lindeijer <bjorn@lindeijer.nl>

* Fix #148

* Close #85

* Close #149

* Fix just one small issue

* Update changelog

* Contain all layer types in an enum

* Implement `as_xlayer` for `LayerType`

* Make id optional

* Rename `ty` to `layer_type`

* Remove `layer_index`

* Move `as_x` functions to tests

* Fix build

* Expose types at the top level

While the modules are great for organizing the code internally, it is more
convenient for users of the create when the types are available at the top
level. This avoids the verbosity like "map::Map".

* Changelog updated

Removed the lines stating where which class has been moved, since all these
classes are not available at the top level again.

Also removed some things which I didn't think we actually noteworthy for people
reading up on what had changed.

* Restored some changelog items

Removal of these items unrelated to how the data types are exposed.

This simplifies code and should be better on memory usage since it is
less fragmented.

Also added `width` and `height` attributes to `Layer`.

Closes #54

Remove Map::source, add Tileset::source, change Image::source to PathBuf

Fixes #72

Move parse functions to its respective types, rename `Colour` for consistency with PropertyValue::ColorValue, add documentation to map members, add Map::source, misc improvements

Closes #99. ldtk seems to output tiled files with no newlines in the csv
data. Tiled itself does always seem to put the newlines in however, it
can open the ldtk files so we should too.

Also add test for parsed image layers.
The test can be also used as an example off how one can get data from
the parsed map.

This commit also fixes a bug that meant the last cell of each row would
not be parsed if the tmx file was base64 encoded.