…idths

Fixes #354

When the header text width exactly matches the maxcolwidths limit, the
existing code added min_padding (2 spaces) on top of the header width to
compute the minimum column width.  This inflated every cell in that column
by 2 extra trailing spaces — visible as unexpectedly wide columns when the
caller had calculated maxcolwidths from the terminal width.

Root cause
----------
minwidths was computed as:

    minwidths = [width_fn(h) + min_padding for h in headers]

then passed as the lower bound to _align_column.  When the maximum data
width (after wrapping to maxcolwidths) equals the header width, the
column ended up being maxcolwidths + min_padding wide instead of the
requested maxcolwidths.

Fix
---
After computing minwidths, for each column that has an explicit
maxcolwidths entry: if the header width is >= the column limit, strip the
+min_padding so the column is exactly max(header_width, maxcolwidth) wide.

When the header is shorter than maxcolwidths the data (wrapped to
maxcolwidth chars) is wider, and the existing behaviour is preserved —
only the case where header_width >= maxcolwidth incorrectly inflated the
column.

Updated tests
-------------
test_wrap_none_value and test_wrap_none_value_with_missingval: these tests
used maxcolwidths=5 with a 5-char header ("Value"); their expected output
encoded the buggy +2 behaviour and have been corrected.

New test
--------
test_maxcolwidth_no_extra_padding_when_header_equals_limit: directly
reproduces the issue example from the bug report.