fix: dataset initialization

[poszu]

Closes #54

Motivation

I looked over the tests in the RandomX C library and I found discrepancies in dataset initialization. The docs say:

In order to use the Dataset, all items from 0 to (randomx_dataset_item_count() - 1) must be initialized.

In effect, all hashes calculated in fast mode (FLAG_FULL_MEM) are invalid.

Changes:

• added a few tests for hashing in both light and fast modes using the vectors borrowed from https://github.com/tevador/RandomX/blob/040f4500a6e79d54d84a668013a94507045e786f/src/tests/tests.cpp#L963-L985;
• fixed RandomXDataset initialization so that all randomx_dataset_item_count() items are initialized;
• changed RandomXDataset::count() into an associated function as it is only a wrapper over randomx_dataset_item_count() and it doesn't require an instance in order to be called.

[hansieodendaal]

utACK

[SWvheerden]

Looks goodf, just remove that dead code

[poszu]

@SWvheerden is there anything else I need to do?

[SWvheerden]

Thanks for the ping, this slipped under the radar.

[SWvheerden]

Is it possible to rebase the commits and sign them with gpg?
I see you singed commit d98bd3a, but not the rest?

The project is blocking me from merging unless all commits are singed.

[poszu]

@SWvheerden

Is it possible to rebase the commits and sign them with gpg? I see you singed commit d98bd3a, but not the rest?

The project is blocking me from merging unless all commits are singed.

Done

[poszu]

@SWvheerden ping