builtin: add dir builtin

[sbinet]

Fixes#12.

[sbinet]

this is a WIP.

@ncw, for dir(), I need to be able to get at the content of the local scope.
how would I do that from within builtin_dir ?

[ncw]

The way I've done that in the past, eg for implementing locals() is to implement them as InternalMethods.

Eg here is where the magic is...

gpython/vm/eval.go

Lines 1578 to 1600 in eaa7d28

	funccallInternal(fn py.Object, args py.Tuple, kwargs py.StringDict, f*py.Frame) (py.Object, error) {
	ifmethod, ok:=fn.(*py.Method); ok {
	switchx:=method.Internal(); x {
	casepy.InternalMethodNone:
	casepy.InternalMethodGlobals:
	returnf.Globals, nil
	casepy.InternalMethodLocals:
	f.FastToLocals()
	returnf.Locals, nil
	casepy.InternalMethodImport:
	returnpy.BuiltinImport(nil, args, kwargs, f.Globals)
	casepy.InternalMethodEval:
	f.FastToLocals()
	returnbuiltinEval(nil, args, kwargs, f.Locals, f.Globals, f.Builtins)
	casepy.InternalMethodExec:
	f.FastToLocals()
	returnbuiltinExec(nil, args, kwargs, f.Locals, f.Globals, f.Builtins)
	default:
	returnnil, py.ExceptionNewf(py.SystemError, "Internal method %v not found", x)
	}
	}
	returnpy.Call(fn, args, kwargs)
	}

This is where the other magic is!

$ git grep InternalMethodLocals
builtin/builtin.go: py.MustNewMethod("locals", py.InternalMethodLocals, 0, locals_doc),
py/method.go: InternalMethodLocals
vm/eval.go: case py.InternalMethodLocals:

That is the only way I could find to avoid using global variables etc. It might be too limiting eventually.

[drew-512]

@sbinet after the py.Context merge, I could look at this out if we think dir() is worth it.

[drew-512]

@sbinet lmk if this is worth adding/improving so we can work towards a cleared PR section. Also, any suggestions on the next version/tag? I think we're worthy of a 0.1 or 1.0 -- open to discussion.

[sbinet]

in order to get the nice discoverability look and feel of CPython into gpython, having a working dir() builtin is definitely a plus.

but as we survived w/o it for almost 4 years, I don't think cleaning up this PR should be a blocker for the next tagged version of gpython.