Adjust coverage resolution based on the number of tests

When running more than 100 tests, the resolution of the coverage output can be confusing:

✗ prop_validates_age failed
   after 1000 tests.
   exactly 18, validated on common day 13% ██▌················· ✓  5%
   exactly 18, validated on leap day    5% █··················· ✓  1%
   close to 18, non-leapling           27% █████▎·············· ✓  5%
 ⚠ close to 18, leapling                1% ···················· ✗  1%
   too young                           30% █████▉·············· ✓ 10%
   not yet born                         5% ▉··················· ✓  1%
   old enough                          70% ██████████████······ ✓ 30%

In this case: ⚠ close to 18, leapling 1% ···················· ✗ 1%

We only covered 0.9% but it is rounded to 1% when displayed.

We should adjust the precision of displayed coverage based on the number of tests, so something like:

-- 100 tests
   close to 18, non-leapling           27% █████▎·············· ✓  5%
 ⚠ close to 18, leapling                0% ···················· ✗  1%

-- 1000 tests
   close to 18, non-leapling         27.4% █████▎·············· ✓  5%
 ⚠ close to 18, leapling              0.9% ···················· ✗  1%

-- 10000 tests
   close to 18, non-leapling        27.39% █████▎·············· ✓  5%
 ⚠ close to 18, leapling             0.09% ···················· ✗  1%

-- 100000 tests
   close to 18, non-leapling       27.392% █████▎·············· ✓  5%
 ⚠ close to 18, leapling            0.009% ···················· ✗  1%

-- etc

Coverage requirements (the right column) can only be integers, at least at the moment. So we might as well keep them with no decimal places.

I suspect this might need to be changed so there is no rounding. It is probably a hangover from before the rounding was done in the reporting code itself.