Changed precision for default localeOptions

[siddhant-nair]

Fixes #82

Was getting wrong results for 1023 and 1001 in default settings due to low precision.

Changed default to the precision of the integer part of the number because 1020 B is a huge round off from 1023 B.

[sindresorhus]

Should this only be done when using the binary option?

[siddhant-nair]

This logic actually wont interfere with non binary numbers in general, since after any 999 value the next value defaults to the next highest unit.

However in a case such as:

// with old code
prettyBytes(827181 * 10e25) // --> 82700000 YB

// with new code
prettyBytes(827181 * 10e25) // --> 82718100 YB

User gets an accurate value

[sindresorhus]

However in a case such as:

Can you add a test case for that?

[siddhant-nair]

There you go