Improve browser focusable tests

[rodneyrehm]

Extracted anything concerning tests/focusable/* from #80.

Changes to the focusable test

• fixing duplicated test element identification
• fixing <input type="reset"> test
• fixing tabbing hangs in firefox by providing defined re-entry point
• fixing tests concerning attribute variations for embedded SVG
• adding tests for SVG documents embedded through <iframe>, <object> and <embed>
• adding test for Shadow DOM (concerning tabsequence)
• adding test for flexbox layout (concerning tabsequence and bug in IE10 and IE11)
• adding test for tabindex ordering in iOS
• adding test for image hiding elements from tabsequence in firefox
• adding test for <form> and <form disabled> (issue in IE)
• adding results for Opera 34

Changes to data tables structure

• merging data-tables/focusable.redirect.html into data-tables/focusable.html, thereby adding the T note to explain "focus target forwarding" and "focus encapsulation".
• moving unused footnotes removal from runtime to build.

Issues with data tables

• meta.expected.json should know the expected redirection target (e.g. for "fieldset legend")
• meta.expected.json needs adjustment for * within <embed tabindex="-1"> and * within <object tabindex="-1">
• focusable.strict.html needs styling for "redirecting" (color, attribute, notes)
• Shadow DOM elements are shown as inert when they should show their correct focusable state
• Add notes for browsers that don't support Shadow DOM
• focusable.strict.html needs improved visualization of the categorical inert in for "SVG document in" [<iframe>, <embed>, <object>] and <iframe> content. (current state is unnecessarily detailed)
• focusable.strict.html needs to check ally.get.focusTarget for inert elements to match the redirecting state (but ignore "parental redirection")
• Shadow DOM host elements are identified tabbable or only-tabbable in Chrome and Firefox, but should probably be a new state along the lines of contained Ignoring this until Tabindex Focus Navigation Explainer led somewhere

Issues with results

• <input> within <form disabled> is inert in IE9-11
• <div style="display: flex"> is focusable in IE10, IE11
• <div id="shadow-host-1"> is identified focusable in IE9-11, but it is not
• <div id="shadow-host-3" tabindex="0"> is identified focusable in IE9-12, but it is tabbable
• <legend> redirects focus differently in Chrome (first focusable in <fieldset>) and Firefox (next item in document's tab-sequence after <legend>)
• <object type="image/svg+xml" data="…" height="0"> is identified as focusable in Safari 9 and iOS 9, but is tabbable?! - fishy
• Image maps are now identified inert or only-tabbable in Firefox - fishy There's definitely something buggy in Firefox

Issues with ally.js functionality

• all platform.name checks should be replaced by platform.layout to increase "derivative browser" compatibility (such as Opera 34, which uses Blink)
• ally.get.focusRedirectTarget should identify focus redirections (not encapsulations)
• ally.get.focusTarget should use ally.get.focusRedirectTarget
• ally.is.focusRelevant should identify elements with ShadowRoot because they might receive focus and ally.is.focusable should revert that
• ally.query.tabbable and ally.query.tabsequence need to compensate swallowed elements (Gecko 1116126) This PR is concerned with improving the focus test coverage and immediate issues with inert/focusable/tabbable/redirecting state
• ally.fix.pointerFocusChildren should also cover <div style="display: flex"> being considered focusable bei IE10 and IE11 While focusing the child of a flexbox-layouted link is a problem, clicking on a div that gets focus is the same as clicking on a scrollable element.
• ally.query.* should accept <object> and <iframe> context and dive into their contentWindows an attempt at this is started in the branch feature/content-document

[rodneyrehm]

Braindump:

---

I'm wondering if the following utilities should rather be exposed in the element namespace, considering they're

element/disabled
element/tabindex-value    (utils/tabindex-value)

element/get-document      (utils/get-document)
element/get-window        (utils/get-window)

element/map/find          (utils/image-map)
element/map/get-img       (utils/image-map)
element/img/get-map       (utils/image-map)
element/label/get-control (get/focus-redirect-target)

I'm not sure about the document/window helpers, they might as well go into the get namespace

get/document      (utils/get-document)
get/window        (utils/get-window)

[rodneyrehm]

The manual browser focus compatibility test has been rewritten from scratch. The following items are left open:

Test anomalies

• manual tabbing test needs to be done slowly, i.e. give the browser some time to actually collect the data in IE9 (likely because of missing requestAnimationFrame)

False results of previous test infrastructure

• area[href].upper only focusable in IE10 - IE13
• iframe and variations are mostly only focusable
• object[src=svg] and variations are mostly inert in IE
• map.object area is focusable in Firefox

Data Tables

• link iframe[src=svg][tabindex=-1] -> file:svg to note 1 and 2
• move results to image map section
• quick/strict refer to querying the dom, not the is/ functions
• mismatch of is/focusable and query/focusableStrict possible!
• shadow dom hosts are now completely inert, but are inert host
• Add note (active) and note for elements that become activeElement because of child content (iframe, ShadowDOM, …)
• verify if <svg> gets host-focus if its content has focus
• Add note for missing :focus