Disable std feature by default in rand_core and rand_jitter

[dhardy]

Implements #702
Resolves #645

Publishing under a new version is the correct way of doing things.

I took the liberty of applying the same logic to rand_jitter, which does have more requirement to use its std feature but is also a "low level" crate (and not yet published, hence not a breaking change).

I also forced rand_jitter to forward the std feature to rand_core, not because it needs it but because it re-exports rand_core and is thus more consistent for users. By the same logic, rand_os now requires std on rand_core.

[vks]

Does the test fail because of the dead links?

[dhardy]

I don't know; I'm looking into the deadlinks now.

[dhardy]

AppVeyor fails since cargo test --package rand_core identifies multiple versions of rand_core — we should use the manifest trick instead. Not sure why it is refering to rand_core 0.3 however. Edit: it's rdrand.

Of course, we also need a compatibility shim for rand_core 0.3 on 0.4, but that has to happen afterwards.

[dhardy]

I can't find the deadlinks when following the Travis CI procedure. If I just do cargo doc --all there are a few, but looks like all are from external dependencies.

I think deploying this will break builds on SGX since that uses RDRAND which depends on rand_core 0.3, so we will need to deploy the compatibility shim immediately. (Probably the only issue though is that sgx.rs uses rand_core from its own dependency, instead of the the version used by rdrand, though since it's not re-exported as rdrand::rand_core getting it right isn't possible now.)

[dhardy]

The Travis deadlinks check is still going on about

Linked file at path /home/travis/build/rust-random/rand/target/doc//index.html does not exist!

many times over. Perhaps there's a missing crate name between the double //, but I'm not sure why it's linking to the index page (I checked all mentions of 'index').

[dhardy]

I was planning on publishing this today, but without further feedback or a solution to that CI error I'd rather hold off a little longer. Any ideas?

[dhardy]

Well, both these CI errors are happening on other PRs, so I will go ahead and merge.