Read `texture` via `_texture` directly in sprite (#3780) ...

* Use `*._texture.*` directly in Sprite.js

It could offer a bit little performance gains.
And  read texture via `_texture` is a common way/style in  PIXI.js

* Update CanvasTinter.js

* Update DisplacementFilter.js

* Update Graphics.js

[v4.4] fix SpriteRenderer for multiple webgl renderers, fix tests, add "withGL" (#3758) ...

* fix SpriteRenderer for multiple webgl renderers, fix tests, add "withGL"

* isWebGLSupported renamed to withGL

fixed lint warnings (#3726)

minor optimisation of quadratic bezier curve calculation. (#3712)

Fast rect blendMode problem (#3707) ...

* fast rect blendMode. Tested.

* add webgl-only test
add test for rgb2hex
fix problem in rgb2hex

add nativeLines option for Graphics (#3328)

Fixed issue where radius of circle may be 0 (#3634) ...

* Fixed issue where radius of circle may be 0

* Update buildCircle.js

Fix for generateCanvasTexture() #3446 (#3605)

Updated link #3368 (#3597)

Improved documentation of the destroy method (#3529) ...

* Improved documentation of the destroy method
Documents the .texture and .baseTexture parameters for the settings object that may not be used for the class itself, but could get propogated down to children
Fixes https://github.com/pixijs/pixi.js/issues/3407

* Improved clarity of docs

Update `clear` for a more thorough cleaning (#3445) ...

* Update `clear` for a more thorough cleaning

Clear something that should be cleared.
`
        this.currentPath = null;
        this._spriteRect = null;
`
if no this patch , some test cases will be failed.


##Test Case
There is some common codes of test cases :
```js
var renderer = new PIXI.WebGLRenderer(900, 600, {
    clearBeforeRender: false,
    backgroundColor: 0xFFFFFF
});
document.body.appendChild(renderer.view);

// the root container.
var stage = new PIXI.Container();
```

### TEST 1
If no  ```currentPath = null ``` ,  the second arc can't be rendered, and there is no error on browser's console.

```js

    var graphics = new PIXI.Graphics();
    stage.addChild(graphics);

    // clear manually
    renderer.clear();

    graphics.clear();
    graphics.beginFill(0xFF0000);
    graphics.arc(150, 160, 100, 0, Math.PI * 2, false);
    graphics.endFill();

   // draw  the first shape
    renderer.render(stage);

    graphics.clear();
    graphics.beginFill(0x0000FF);
    graphics.arc(150+50, 160+50, 100, 0, Math.PI * 2, false);
    graphics.endFill();

   // draw  the second shape
    renderer.render(stage);
```

### TEST 2
If no  ```_spriteRect = null ``` ,  the first rect can't be rendered, and there is no error on browser's console.

```
    var graphics = new PIXI.Graphics();
    stage.addChild(graphics);

    // clear manually
    renderer.clear();

    graphics.clear();
    graphics.beginFill(0xFF0000);
    graphics.drawRect(150, 160, 100, 60);
    graphics.endFill();

    // draw the first shape
    graphics.updateTransform();
    graphics.renderWebGL(renderer);

    graphics.clear();
    graphics.beginFill(0x0000FF);
    graphics.drawRect(150+50, 160+50, 100, 60);
    graphics.endFill();

    // draw the second shape
    graphics.updateTransform();
    graphics.renderWebGL(renderer);

    // flush manually
    renderer.flush();
```

### Summary

If don't call  `graphics.clear();` before draw the second shape , there will be another error.

If add 
```
        this.currentPath = null;
        this._spriteRect = null;
```
in `clear()` everything will be OK, and I test many other examples, this patch doesn't break anything else, it's safe.


I know when we want to draw multi shapes , we could use multi graphics objects.
But we shouldn't prevent users to use one shared graphics to draw multi shapes.

* Remove trailing spaces.

Test case & fix for Graphics._calculateBounds calling updateLocalBounds every time (#3444) ...

* Test case for #3414

* Fixes #3414

* fixed expectation

Graphics - getLocalBounds after generateCanvasTexture (#3398) ...

* added graphics.getLocalBounds expectation

* fixes #3274

add comments and change test description

fix/graphics: get points for newly created shape (#3433)

Merge pull request #3290 from pixijs/master ...

Merge master -> dev

Merge pull request #3206 from pixijs/dev-vao-state-cache ...

Dev vao state cache

spelling

Cache VAO states

Fix anticlockwise sweep calculation (#3249) ...

* Fix anticlockwise sweep calculation

* Simplify sweep calculation

* Sweep typo

Merge branch 'master' into dev

minor eslint fixes

get rid of trigonometric stuff. ...

Explanation: sinTheta*w is actually dy, cosTheta*w is actually dx.

switch to arctan2

needs absolute value

fixes #3168

lint fixes

fixes #2824

Fixed variable naming issue reported in #2903 (#3035)

updated some doc comments (#3038)