fread can underestimate nrow on rerun when type bump and healing with fill=TRUE

[ben-schwen]

Closes #5110

Still need to come up with a clever test for this.

[github-actions]

• HEAD=fread_reallocate stopped early for DT[by,verbose=TRUE] improved in #6296
  

Generated via commit 51057cd

Download link for the artifact containing the test results: ↓ atime-results.zip

Task	Duration
R setup and installing dependencies	5 minutes and 14 seconds
Installing different package versions	10 minutes and 10 seconds
Running and plotting the test cases	2 minutes and 39 seconds

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.13%. Comparing base (67db7f7) to head (51057cd).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #7385   +/-   ##
=======================================
  Coverage   99.12%   99.13%           
=======================================
  Files          85       85           
  Lines       16637    16556   -81     
=======================================
- Hits        16492    16412   -80     
+ Misses        145      144    -1     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.