Trigger the call of pdfjs on MIME type rather than extension

[babolivier]

I just noticed that pdfjs was systematically called on any URL ending with .pdf, such as http://sdfgsdfgsd.com/test.pdf (which doesn't exist) or https://github.com/Intel-Media-SDK/samples/blob/master/samples/sample_decode/readme-decode_linux.pdf (which returns HTML content, not readable by a PDF reader). I think that pdfsj should be called regarding the response MIME type (application/pdf for PDF) and not how the URL is ending.

• Platform (Win7, 8, 10? macOS? Linux distro?): Linux
• Brave Version: 0.11.1
• Steps to reproduce:
  1. Open a new tab
  2. Write or paste an URL (leading to an existing resource or not) containing a domain name and a path ending in .pdf

[diracdeltas]

++, i think this is causing problem on PDFs that redirect to an HTML page for auth (ex: in slack)

this may be fixed by taking out the setPDFLocation calls in js/stores/windowStore.js, which are a workaround for PDFs not loading at all unless they fire an Electron navigation event

[diracdeltas]

verified that pdfjs does MIME detection in webRequest.onHeadersReceived so this should be fixed by reverting the workaround for the other bug

[luixxiul]

in 0.11.2?

[bbondy]

we will put all in 0.11.2 that's in master and freeze as soon as chromium 52 is avail.