`vstack` does not preserve dtype of `Quantity` columns

Description

When applying vstack on two or more QTable objects which have an integer Quantity column, I find that the dtype of that column is unexpectedly converted from int to float.

Minimum example:

>>> from astropy.table import QTable, vstack
>>> from astropy.units import Quantity
>>> tbl = QTable({'x': Quantity([1], dtype=int)})
>>> vstack((tbl, tbl))['x'].dtype
dtype('float64')

I expected to see dtype('int64') above.

The error does not occur when Table is used instead of QTable, or when Column is used instead of Quantity.

Debugging attempt

My debugger shows that the following line in _vstack() returns a Quantity with the incorrect dtype:

https://github.com/astropy/astropy/blob/10d27d8292bafb3ab502f4147dd707473e69af71/astropy/table/operations.py#L1288

…unfortunately I wasn’t able to understand the DataInfo mechanism sufficiently to understand the origin of this error.

System Details

I am using today’s master branch, i.e.

Python 3.8.6 (default, Oct  8 2020, 14:06:32)
[Clang 12.0.0 (clang-1200.0.32.2)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import astropy
>>> print(astropy.__version__)
4.3.dev20+g10d27d829

[barentsen]

@pllim Unfortunately it won’t! I just checked out #10950 and it does not fix the bug/behavior reported here. I believe this is a separate issue.

[barentsen]

(I’m closing this issue because I’m now convinced that we shouldn’t be storing integers in Quantity objects in the first place. Please reopen if anyone disagrees!)