Consider removing .with_multisampling(4) from glium examples

[Type1J]

I'm trying out Conrod on a Pinebook Pro (Rock64 SoC: ARM64 CPU with Mali GPU). The "Panfrost" open source driver for Mali doesn't seem to support multisampling pixel formats, and so it fails to start:

thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: GlutinCreationError(NoAvailablePixelFormat)', backends/conrod_glium/examples/canvas.rs:26:19

This can easily be fixed by removing the requirement of multisampling (the call to .with_multisampling(4) just above that line in the same file.

I was able to interpret the error message to see what possibly needed to be done, but someone trying out Conrod for the first time may see this as a problem with Conrod and abandon it. I know that multisampling may make it look prettier, but it's not that much prettier, and I don't feel that showing the examples looking their best (when the second best is so much similar) is worth turning off potential users. There could be 2 versions of the same example with 1 using multisampling just to show it off, but labeling it that way gives the expectation that if it doesn't run, then it's the hardware's (or driver's) fault, and not Conrod's.

[Type1J]

An alternative would be to just handle the error, and retry.