* Consistent getPoints and getSpacedPoints on ellipse

They work as getSpacedPoints used to (but with the final point)
The arc between aStartAngle and aEndAngle is always forced in the 0, 2
PI area, as getSpacedPoints did (but not getPoints)

* Refactor Path class

- got rid of .actions
- getPoints() moved to CurvePath
- toShapes() moved to new class ShapePath

* Path: simplify usage of this.currentPoint

* Optimize LineCurve.getPoint(1) as suggested by @rfm1201 in #8952

- this minimizes some floating point errors when t = 1

* Fix up missing changes to CurvePath and EllipseCurve

- as commented by @rfm1201
- using f040453 as reference

* Path: update currentPoint for .splineThru(), add automatic lineTo()

- for ellipse

* CurvePath: address some removeDuplicatesInAPath() concerns

- also cleanup some old reference to .actions

* Path/CurvePath getPoints() shouldn't return sequence of duplicate points

* In path.absellipse, lineTo is drawn only to connect existing path

- when necessary

In the actions array of the Path, the last elements of the args array
need to be the x and y of the last point of the segment added to the
path.
Not the case with splineThru => cause an exception if you add an other
segment after the pline (without a moveTo).
Used the same trick as absellipse does.

Changed
if (aRotation ) {
to
if ( aRotation !== 0 ) {

Adds an optional rotation parameter to THREE.Path.ellipse and THREE.Path.absellipse.

eg. / e.g.
vertice / vertex
fullframe / full-frame
todo / TODO
unreckognized / unrecognized
ontop / in front
initialise / initialize
unforms / uniforms
overriden / overridden
interpolatation / interpolation
subsurve / sub-curve
Catmull-rom / Catmull-Rom
consruct / construct
colinear / collinear
perpenicular / perpendicular
minimim / minimum
segements / segments
speficied / specified
compatable / compatible
egde / edge
initalization / initialization
shiftet / shifted
childrens' / children's
ambigious / ambiguous
barycoordinates / barycentric coordinates
equi distance / equidistant
equidistance / equidistant
ist / is

And remove one `console.log`

Fixes #5570

THREE.Path.prototype.toShapes: 1) Bugfix: if all subpaths have the same wrong winding order; 2) new optional parameter: noHoles

THREE.Path.toShapes is now independent of the order of specification of holes and shapes. It moves holes to those shapes that contain them if unambiguous. Otherwise it respects specification order.

- aClockwise goes anti-clockwise
- The end point is now correct and no longer breaks path continuity

- Path.js (esp. Arcs) is still confusing, but as of now
- Ellipses should work
- Path still uses "PATH.ACTIONS"
- the internal curve representation doesn't seem to be used yet
- Path.arc() uses .absarc() which uses .absellipse()
- examples/webgl_geometry_shapes.html should run now

removing less necessary stuff from Path.js. May still have to decide whether to migrate to using .curves from .actions