Describe the issue:

I am using np.format_float_positional() to output exactly 3 digits of precision.
See the code example below.
The output is:

0.0000314 0.000314 0.00314 0.0314 0.314 3.14 31.4 314. 3140. 31400.
0.0000310 0.00031 0.00310 0.0310 0.310 3.10 31.0 310. 3100. 31000.
0.0000250 0.000250 0.00250 0.0250 0.25 2.50 25.0 250. 2500. 25000.

In the second line, 0.00031 should be 0.000310.
And in the third line, 0.25 should be 0.250.

Reproduce the code example:

import numpy as np

for base_value in [np.pi, 3.1, 2.5]:
  print(' '.join(np.format_float_positional(value, fractional=False,
                                            precision=3, unique=False)
                 for value in base_value * 10.0**np.arange(-5, 5)))

Error message:

No response

NumPy/Python version information:

1.21.6 3.8.10 (default, Mar 15 2022, 12:22:08)
[GCC 9.4.0]