Implement no_std support

[philipc]

Fixes #90

I also changed the travis builds so there's not quite so many builds.

This still needs a gimli update, so the nightly travis build will fail.

[philipc]

@roblabla Do you expect any more gimli changes will be needed for the unwind support, or should we do a minor gimli release now?

[roblabla]

CC @main-- .

The only other thing unwind-rs' fork had was a bandaid that would parse DW_CFA_GNU_args_size, and ignore this result. From what I read, that implementation is wrong, and anyways is only necessary for i386 windows binaries (source: https://reviews.llvm.org/D38680)

I started a proper implementation for this. I guess we can wait for that before doing a gimli minor bump. I'll try to get a PR up during the week-end.

[main--]

From what I read, that implementation is wrong, and anyways is only necessary for i386 windows binaries

Very interesting, most certainly not how I encountered it but I'm unable to reproduce it so ¯\(ツ)/¯

I started a proper implementation for this. I guess we can wait for that before doing a gimli minor bump. I'll try to get a PR up during the week-end.

Sounds good.

[coveralls]

Coverage remained the same at 75.882% when pulling 48f5eec on philipc:no_std into b1ac5ae on gimli-rs:master.

[philipc]

This should be ready for review/merge now.