FMSLogo Code

Use wxSingleInstanceChecker to detect if another instance of FMSLogo is running on startup.
This prevents multiple instances of FMSLogo from running on GNU/Linux unless to they were given a specific program to run.
The CreateMutex logic on Windows was preserved until it can be confirmed that wxSingleInstanceChecker can grab the exact same mutex.
Otherwise, FMSLogo 8.5.0 couldn't detect if FMSLogo 8.4.0, the installer, or the uninstaller is running, and vice versa.

Refactoring: Rename NODETYPES to NODETYPE.
When the type was an "unsigned short", a plural made sense because it was the set of all possible types.
As an enum, a singular makes more sense, because it's the type of the associated NODE.

Change NODETYPES from "unsigned short" to an enum.
This adds a little more type safety and makes debugging easier (the debugger shows the symbolic name, not the numeric value).
While this increased the size of the type from 2 to 4 in gcc, it didn't increase the size of the NODE.
If the new size of NODETYPES proves to be problematic, it can be changed to an "enum : unsigned short".

Fix Bug #598; QUEUE and DEQUEUE now behave correctly when the queue name is shadowed by one of their own local variables.
The fix is to rewrite these procedures as macros that get evaluated in the caller's context.
For performance, when the queue name doesn't clash with a local variable, the previous implementation is run.

Update the tests so that a full test pass can run on GNU/Linux.
In some cases, this meant finding ways to make tests run differently on Windows or GNU/Linux.
In most cases, this meant not running tests whose execution depends on Windows,
such as any test that uses UIAUTOMATIONLIB, which uses DLLCALL for win32 API.

Refactoring: Replace CURRENTWORKINGDIRECTORY with GETCURRENTDIRECTORY in dllcall.
This removes duplicated code.

Adjust some of the assert statements that I had added to the inline functions.
The getstr* functions now call is_word() which works for all string types.
Confusingly, is_string() is a special case of is_word().

Also treepair__tree is not always called on a tree.
This may be a bug, which I'll look into later.
For now, it's more important for fmslogo to run with asserts than to have this assert.

Fix a typo in a comment

Comment out the "assert(is_tree(tree))" that I recently added to settreepair__tree(tree, treepair).
There are code paths, notably in treeify_line, which invoke untreeify on nodes that aren't trees.
That could be a bug, but having the assert is counter-productive as it currently fires on nearly any procedure call.
If the untreeify logic can be rewritten to only be invoked on trees, then I will re-enable the assert.

Add comments to explain the purpose and function of the two different stacks used by the evaluator.