* Revert service worker usage of MessageChannel
Reverts 66b6adf035
While I'm not 100% about this, I think 66b6adf035 causes resourses to occasionally not load. I believe this can happen if the service worker is unitilized while the webview remains active. I can't reproduce this myself so it may be related to memory pressure or resource usage, however relying on the service worker not being reinitilized does seem like a potentially bad idea https://stackoverflow.com/questions/34775105/what-causes-the-global-context-of-a-service-worker-to-be-reset
Will investigate if there's another way to achive this since using MessagePort did clean up the code and slightly improve performance
* Bump webview commit versions
* Switch webview service-worker to use message channel
This change hooks the service worker used for loading webview resources directly up to the main VS Code process over a message channel. Previously this communication had to go through an extra hop through the webview
This simplifies the logic somewhat (although this change required adding extra logic to exchange the message port). It also improves performance a little
* Update webview content commit version
* Enable webview tests on web
These tests pass fine on Chrome and webkit. On Firefox, the two about loading resources currently fail
I spent a while trying to debug what is going on with these tests. It seems like the service workers is registered properly but `fetch` is never called on it. Skipping these for now
* Remove only
* Fix test
* first cut upload log files from integration test runs
* always publish log files
* move logs into integration-tests folder
* also produce logs for smoke tests
* store remote log files
* wire in logspath arg properly
* always publish logs
* enable logs for browser based smoke tests
* log only in verbose
* linux - remove `--no-sandbox` flag
We should not disable sandbox on Linux.
We only added this to support running on Kernel <= 3.8 where setuid sandbox is used.
* fix ci
* add some docs