API to know if the square is repairable

[Wondertan]

Currently, users must count and assume that 1/4 of shares is enough to start repairing. Instead, a more precise API should be provided to tell the user if the square is repairable. This way, we would:

• Be able to measure the exact time it takes to repair the square
  • Repair method, currently, does repairing for rows and cols silently on each call even when an error happens
  • This spreads repairing over multiple tries with time gaps based between each try
  • The gap is based on another network share being retrieved which kicks off another try
  • Such gaps pollute the time results of each Repair try

(no offense to solveCrossword method, really love how its written, besides solveCrosswordRow/Col being split instead of general code path)

• Lower the number of idle repairing tries

[adlerjohn]

So basically this function would assume that you can fully reconstruct any rows/cols with enough shares, then keep making this assumption until either the square is assumed fully repaired or it runs out of at-least-half-full rows/cols? Removing the overhead of tracking Merkle roots and mismatches, it should be much faster to execute (and could even be executed iteratively if it returns the set of shares that would be repaired so new indices can be inserted).