Patch for #39 Sidechannel resistence of uECC_sign disabled

[winnietwo]

I removed the g_rng_function variable from ecc_dh.c, as there is another one in ecc.c.
I access the g_rng_function variable in ecc.c by the uECC_get_rng(). In this way, side-channel resistance should be enabled again.

[mczraf]

This patch effectively enables the provided side-channel countermeasure in the EC-DSA sign procedure. Thanks @winnietwo for your contribution to enhance TinyCrypt!

I have noticed that the same problem preventing EC-DSA sign to use the side-channel countermeasure was replicated in the EC-DH algorithm. I have fixed the EC-DH implementation in the commit above using the same strategy described by @winnietwo for EC-DSA.

[mczraf]

This patch effectively enables the provided side-channel countermeasure for EC-DSA sign.

[ceolin]

I'm confuse here. This change just has impact if one sets its cryptographically-secure function using uECC_set_rng() and this is different from default_CSPRNG, right ? Is there a reason to have different random generators ? Anyway, the patch itself is good, it makes the code cleaner.

[winnietwo]

Is there a reason to have different random generators?
The default random number generator is for Linux. On many non-Linux embedded devices, it's necessary to have your own implementation.