[easy] Fix lower bound of zk rows for chunking

[querolita]

When chunking is in place, the number of rows reserved for zero knowledge must be augmented accordingly. In particular,

zk_rows > (2 * permuts * chunks - 2) / 7

that means that zk_rows must at least be (2 * permuts * chunks - 2) / 7 + 1 or what's equivalent, (2 * permuts * chunks - 5) / 7. Examples of this are here.

Nonetheless, on the Kimchi side, the strict lower bound (2 * permuts * chunks - 2) / 7 is created here and is used incorrectly when triggering the error message ProverError::NotZeroKnowledge. Meaning, it is raised whenever the number of rows is smaller than that value, when it should be raised anytime it is smaller or equal (because by definition, the number of rows must be larger than that value).

This PR fixes that mismatch and adapts the error message accordingly.

Example:

With 2 chunks, the message used to read there are not enough random rows to achieve zero-knowledge (expected: 4, got: 3) whereas with 2 chunks you need at least 5 rows for zero knowledge.

Partially addresses o1-labs/o1js#1806

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 69.54%. Comparing base (0583e99) to head (14c7e1f).
Report is 63 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2590      +/-   ##
==========================================
- Coverage   70.95%   69.54%   -1.42%     
==========================================
  Files         240      244       +4     
  Lines       57669    57644      -25     
==========================================
- Hits        40919    40088     -831     
- Misses      16750    17556     +806     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.