Add tests for no-std

[elichai]

This will not compile if std is somehow involved.
You can see yourself by removing default-features = false from the Cargo.toml and you'll get:

error[E0152]: duplicate lang item found: `eh_personality`.
  --> src/main.rs:37:1
   |
37 | pub extern "C" fn rust_eh_personality() {}
   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = note: first defined in crate `panic_unwind` (which `std` depends on).

error[E0152]: duplicate lang item found: `panic_impl`.
  --> src/main.rs:83:1
   |
83 | / fn panic(info: &PanicInfo) -> ! {
84 | |     unsafe { libc::printf("shi1\n\0".as_ptr() as _) };
85 | |     let msg = info.message().unwrap();
86 | |     let mut buf = Print::new();
...  |
89 | |     unsafe { intrinsics::abort() }
90 | | }
   | |_^
   |
   = note: first defined in crate `std` (which `secp256k1` depends on).

error: aborting due to 2 previous errors

This currently only runs in release mode due to a bug that debug mode always tries to link to libunwind (which isn't available) even when we use abort instead of panic.

[apoelstra]

Can you clarify the motivation for this PR?

[elichai]

@apoelstra Yes, #172.
This PR made me doubt our no-std support for a second so I wanted to write my own simple check to make sure that I'm not missing anything

[jonasnick]

This is very nice. Approach ACK. I think it would be helpful to add a few sentences of documention (by adding a header main.rs?) mentioning what this is (a minimal nostd example that exercises some secp functions and features) and the kinds of error messages expected when something accidentally pulls in std.

[apoelstra]

utACK. Looks great.