Rework test to restore 100% test-coverage

Previously we invoked our parser, to parse the input, in our

`Evaluate` call.  This meant that evaluating the same script
multiple times wasn't possible:

* First run:
  * Evaluate parses the program, executes each command.
* Second run
  * Evaluate has a parser which has already consumed the input.
  * There is nothing to parse.
  * So nothing is executed.

This pull-request moves the parser-consumption to the constructor,
which necessitated changing the constructor's signature, so that
the parser will always consume all input ahead of the call to
`Evaluate`.

The end result is you can run:

* New(..)
* Evalute
  * Runs the program.
* Evaluate
  * Runs the program, again.
* ..

This closes #25.

Removed local module override

Added an example of embedding the scripting language. (#24)

This closes #21, by showing how to embed the TCL interpretter inside
an application of the users:

* Loading scripts.
* Exporting functions from the host application to the TCL environment
* Checking return-values & etc.

Revert the tokenization of maths-primitives; it broke comments & etc

Fixed issue where "$a+$b" failed.

Since spaces were required.

Added -version flag