crypto.ecdsa: fix handling sign with custom_hash

[blackshirt]

This small PR following previously already merged PR in fix-no-hash commit.
The current handling with custom hash was still contains some bogus thing without additional test (sorry, its my bad).
Its resolved in the current PR, by the way with:

• call .reset() on the custom hash before writing messages
• previously, its not call .write() before .sum() ..its does not matching to expected result, when compared, for examples digest produced with sha256.sum256(message) directly. The equivalent of sum(message) was

_ := custom_hash.write(message)!
digest := custom_hash.sum([]u8{})

• Add some test to verify expected behaviour

Thanks ...

[huly-for-github]

Connected to Huly®: V_0.6-22043