`tabulate.tabulate([['126,000']], tablefmt="html")` raises error starting in 0.8.8

Starting in 0.8.8 we’re seeing errors when converting to HTML with large numbers encoded as strings:

import tabualate
tabulate.tabulate([['126,000']], tablefmt="html")

Error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-2-37ea0f4d62eb> in <module>()
      1 import tabulate
      2 print(tabulate.__version__)
----> 3 tabulate.tabulate([['126,000']], tablefmt="html")

3 frames
/usr/local/lib/python3.6/dist-packages/tabulate.py in _format(val, valtype, floatfmt, missingval, has_invisible)
   1002             return val.replace(raw_val, formatted_val)
   1003         else:
-> 1004             return format(float(val), floatfmt)
   1005     else:
   1006         return "{0}".format(val)

ValueError: could not convert string to float: '126,000'

This works in previous versions of tabulate, see https://colab.research.google.com/gist/tomhennigan/55ffecb257911296cf181ae968a7356a/tabulate-0-8-8-bug.ipynb for an example.

[astanin]

For now I added tabulate([['126,000']]) to the list of regression tests and reverted changes which caused this problem (pr #111).

[astanin]

This bug is due to a feature introduced in 0.8.8 (pull request #51). A regression test was added, and the feature was disabled in 0.8.9 (pull request #111). It will be implemented differently.