Expose seed argument to hash_values

[ayushdg]

Description

This PR exposes the seed param to hash_values that is already supported by libcudf's hash method.
Closes #12775

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[brandon-b-miller]

If the seed has no effect, I think maybe we should warn, at least since we have some flexibility here given this has no equivalent pandas API.

[ayushdg]

Another alternative is to separate out the test for warning into a separate pytest and not use seed at all for this one.

[ayushdg]

One observation while testing out hash_values for murmur3 with different seed values is that there seems to be little variance and somewhat of a linear relationship b/w the seed and the hashed values. I'm not sure if this is expected behavior or not, but most implementations of murmur3 from 3rd party libraries seem to give much more varied results.

ser = cudf.Series(["hello world"]).str.character_ngrams(5,False)

ser.hash_values(seed=0)
ser.hash_values(seed=0xFFFFFFF)

|               | Seed=0     | Seed=1     | Seed=0xFFFFFFF |
|---------------|------------|------------|----------------|
| 0             | 3267589120 | 3267589185 | 3376188480     |
| 1             | 1877329862 | 1877329927 | 2082093178     |
| 2             | 1758410039 | 1758410102 | 1664142089     |
| 3             | 1327917612 | 1327917677 | 1557763604     |
| 4             | 2896800200 | 2896800265 | 3210106488     |
| 5             | 197501035  | 197501098  | 3825621        |
| 6             | 2580395700 | 2580395765 | 2452769164     |
| dtype: uint32 |            |            |                |

[brandon-b-miller]

Approving. Not sure about the variance question but given that we're just wrapping here I'd say it's a libcudf question.

[ayushdg]

/merge