Requesting faster arbitrary-precision Ln and Power implementations

[peteroupc]

The current version of Ln and Power in RadixMath.cs (Ln, Power) is slow and full of accuracy "hacks" (an example showing how slow it is, is Ln for a 1000-decimal digit integer). I would like to replace Ln and Power with a simpler and faster version with the following requirements:

• Works for arbitrary-precision decimal and binary floating-point numbers.
• Ensures accuracy to 1/2 unit in the last place for up to 36 digits EDecimal precision or 64 bits EFloat precision, and ideally to 1 unit in the last place for other precisions.
• Implements the General Decimal Arithmetic Specification just like the current version.
• In the public domain (since this is a Creative Commons Zero or CC0 library).