Increase digits printed by plot_bvec

[jaemolihm]

seedname.bvec file is used by EPW, and I found small but nonzero (~1E-6) relative error from the round-off errors in bvec. Such a small error rarely affects physics but it makes debugging tricky. So I believe it is worth changing.

In addition, testw90_bvec did not test the bvec data (it only checked the sizes). I updated the parser to check bvec data and updated the reference data.

[hjunlee]

@jaemolihm I already changed the number of digits here (#337) and for the moment, I don't want further changes.

In addition, this part is only related to EPW and I would like to ask all people who want to do MRs related to EPW in Wannier90 to directly contact me, the lead developer of EPW, before doing MRs.

[jaemolihm]

@hjunlee This PR is a change to Wannier90 source code, not EPW source code. This code is in the Wannier90 upstream. Thus, I believe the code should be managed in the same way as other parts of Wannier90, i.e. as an openly-developed community code. I have followed the Wannier90 contributors guide and opened a Pull Requests.

(I think the situation is different from the case of EPW in QE, where the whole EPW/ folder is managed by the EPW team. For changes to the EPW source code, I will certainly follow your instruction.)

In addition, it is not correct that this part (subroutine plot_bvec) is related only to EPW. One can write the bvec file using the write_bvec flag, whether or not one is using EPW. I have faced problems using the seedname.bvec file for my own program (not EPW), thus this PR.

If the reason for objection is that such a change in printed digits is only a quick-and-dirty fix and one should aim for a more fundamental change (e.g. binary IO or directly accessing the variable via library mode), I would accept the argument and close this PR.

[jaemolihm]

@hjunlee In the future, I will make sure I contact and notify you when I make a PR that would impact the EPW code.

[giovannipizzi]

Thanks @jaemolihm! This change is backward compatible and should not affect/break the behaviour of external codes, and I think it's an important improvement (and also comes with tests - thanks!), so I am going to merge this.
If this creates problems, we can discuss in a new issue.