Fix intcomma returning scientific notation for large floats - #432

Fixes #431

Changes proposed in this pull request:

• When a float's string form uses an exponent (1e16 and up, or very small values like 1e-05), format it through Decimal(str(value)) so it comes out as a plain grouped number instead of 1e+16
• Keep the trailing .0 so the output matches what smaller floats already return (intcomma(1e15) is 1,000,000,000,000,000.0)
• Ints, floats that don't use an exponent, and the ndigits path are unchanged
• Added cases to test_intcomma plus one de_DE check in test_i18n

Full test suite (with translation binaries compiled), prek and mypy pass locally.