RoundUp doesn't error for large negative number of digits

Two problems, with decimal

> RoundUp( 123.12345678919, -28 )
10000000000000000000000000000

> RoundUp( 123.12345678919, -29 )
0

We should generate a runtime overflow error for the second case, consistent with #NUM! in Excel.

For floating point

> RoundUp( Float(123.12345678919), -14 )
100000000000000

> RoundUp( Float(123.12345678919), -15 )
999999999999999.9

> RoundUp( Float(123.12345678919), -16 )
123.12345678919

> RoundUp( Float(123.12345678919), -300 )
123.12345678919

> RoundUp( Float(123.12345678919), -400 )
123.12345678919

Not at all consistent with Excel, all numbers should be 1 with a variable number of zeros following. -14, -15, and -16 is an interesting sequence. For -400, we should generate an error.