Merge pull request #8 from kdsch/nayuki

README.md: add fast fibonacci algorithm example

README.md: add fast fibonacci algorithm example

Merge pull request #7 from kdsch/6-move-benchmarks

move benchmarks into tests

testdata/test.kp: change benchmark case

fix benchmark function

move benchmarks into tests

sloc: use find -exec, not xargs

README.md: fix list formatting

README.md: rewrite

refinements

go fmt ./...

remove unused file; rename a file

add license

move all files up one directory

change import path, become a module

progress!

We're in the middle of some changes. Stay tuned.

improve performance

Linked lists as expressions seem to speed up the machine;
particularly, shifting to the next word is much faster than ropes.
Still unclear whether this is true for stacks; should investigate
more.

add variadic bindings

This allows a single bindv expression to bind multiple values
from the stack to a sequence of variables.

let there be lists

Any decent programming language has data structures. We had Expr
and Stack, but now both have been merged into Seq. This is all
for the better. It gives us much flexibility and power. Furthermore,
we've introduce [ ] syntax for lists, backed up by Seq.

We had to abandon the linked list that was the sole implementation
of Stack. It will be missed. But std.Rope is much more powerful;
it works quite well as either a stack or an expression. It's a good
Seq.

We've introduced new builtins for lists.

One wrinkle introduced: lists can contain arbitrary expressions,
which are reduced when the list is reduced. These feels somewhat
like calling a block; but instead of the block working with the
main stack, it works on a separate stack.

This means that the interpreter embraces a self-replicating machine
model, which is not physically realistic. I have not determined
whether I care about this. In the long term, we should maybe not do
this.

lexer: gofmt