Use the c-side constructor of numpy's dtype to expose the typobj attibute

[kif]

close #1436

@s-sajid-ali could you test this branch ?

[codecov]

Codecov Report

Merging #1439 into master will decrease coverage by 0.11%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1439      +/-   ##
==========================================
- Coverage   84.94%   84.82%   -0.12%     
==========================================
  Files          17       17              
  Lines        2105     2128      +23     
==========================================
+ Hits         1788     1805      +17     
- Misses        317      323       +6

Impacted Files	Coverage Δ
h5py/__init__.py	60.27% <0%> (+6.27%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 273dfa9...8f253d7. Read the comment docs.

[takluyver]

Close/reopen to rerun tests.

[takluyver]

I can't reproduce the error either. :-/

[takluyver]

Apparently @s-sajid-ali can't reproduce this either. @kif do you think the change is an improvement anyway, or should we close it?

[scopatz]

This change is probably an improvement as it uses the CAPI

[kif]

Using the C-api is not always a good move. In this case, apparently it changes little. I am pretty agnostic. if @s-sajid-ali cannot reproduce the bug, I would probably drop the PR as it solves no issue.

[kif]

We can leave it open for a month or so in case this bug pops-out again then.

[takluyver]

I'm inclined to agree with 'if it ain't broke, don't fix it'. Good idea to leave it open and obvious for a bit in case someone else can reproduce it, though.

[scopatz]

This still seems like a positive change to me