Un-ignore test-suite runtime-styling tests

[lucaswoj]

fixes #2797

cc @jfirebaugh

ref mapbox/mapbox-gl-test-suite#123

Todo

• figure out if it is ok to not destroy the gl context in test-suite
• fix [Error: Invalid file signature] log messages due to lack of JPEG support
• fix CI failures due to wait operations returning too early
• fix CI segfaults

[lucaswoj]

Remaining test suite failures appear to be related to the implementation of wait. There are some cases in which wait will return despite some resources being pending. Proper resolution may require finishing #1715

For reference, this patch "fixes" the test suite

diff --git a/test/suite_implementation.js b/test/suite_implementation.js
index ef42c28..56474dd 100644
--- a/test/suite_implementation.js
+++ b/test/suite_implementation.js
@@ -90,7 +90,9 @@ function applyOperations(map, operations, callback) {
     } else if (operation[0] === 'wait') {
         var wait = function() {
             if (map.loaded()) {
-                applyOperations(map, operations.slice(1), callback);
+                setTimeout(function() {
+                    applyOperations(map, operations.slice(1), callback);
+                }, 500);
             } else {
                 map.once('render', wait);
             }

[lucaswoj]

This is ready to go on 🍏

How does this look @mollymerp @jfirebaugh @mourner?

[jfirebaugh]

When would gl be falsy here? Instead of adding a guard clause, can we eliminate that case?

[lucaswoj]

A falsy value for gl indicates that we have manually destroyed the GL context (see this code)

Tracking this information prevents "invalid GL context" errors in the runtime styling tests when asynchronous operations are executed after the context has been destroyed.

I will investigate some alternate designs today.

[lucaswoj]

Implemented a (hopefully) cleaner solution in ff07754

[lucaswoj]

Updated and ready for 👀

[jfirebaugh]

🚢

[lucaswoj]

Before shipping this PR, I am investigating an increased rate of intermittent test-suite segfaults that only occur on CI and seem to be a result of this PR.

[lucaswoj]

The CI segfaults are annoying but should not block this release. I will create a separate issue to track their resolution.