Summary of changes to SoftFloat

August, 2018: Package updated to SoftFloat-3e; Hercules modifications
        updated by Stephen Orso.

    Summary of material changes between SoftFloat-3a and -3e

    General changes:
    - Support added to specify value returned on over/underflow
      result or NaN input when converting to (un)signed integer,
      either 32-bit or 64-bit.

    Changes to specific functions:
    f32_scaledResult.c - corrected long-standing error that caused
        large subnormal values to be treated as normal.
    s_roundPackToI32.c renamed to s_roundToI32.c
    s_roundPackToI64.c renamed to s_roundToI64.c
    s_roundPackToUI32.c renamed to s_roundToUI32.c
    s_roundPackToUI64.c renamed to s_roundToUI64.c
    s_subMagsF32: re-written with a fast-path added for subtraction
        of values having the same binary exponent.
    s_subMagsF64: re-written with a fast-path added for subtraction
        of values having the same binary exponent.

    Summary of changes to Hercules modifications

    source/8086-SSE/specialize.h
    s_addMagsF128.c
    s_addMagsF32.c
    s_addMagsF64.c
    s_roundPackToF32 - Accomodate code reordering in SoftFloat-3e.
    s_roundPackToF64 - Accomodate code reordering in SoftFloat-3e..
    s_roundPackToUI32.c - Increased rounding bits from 7 to 12
    s_subMagsF32.c
    s_subMagsF64.c

    Hercules modifications no longer needed for SoftFloat-3e.
    - f64_to_ui64

