v0.3.2

This contains huge acceleration for division:

• if the dividend fit in one digit, use a specialized version of a long division
• otherwise, use the Burnikel-Ziegler recursion

v0.3.1

Fix a corner case in Newton-Raphson convergence check

v0.3.0

• faster division algorithm (newton-raphson)
• adds import/exports to files
• add an example program that computes bits of sqrt(2)
• adds copy_from API to copy an integer into an existing one (preserving the allocation).
• Fixes an issue in shift operations that used to cause a reallocation

v0.2.8

• fix compilation for apple-darwin-x86_64
• introduce an API rsub_assign similar to sub_assign reversed to avoid unnecessary allocation
• fix random generation's very least significant numbers in some cases
• avoid unnecessary reallocations in add_assign
• introduce an API set_to_mul to reuse an existing integer's allocation to be the result of a multiplication operation
• introduce a (private for now) new integer type BigFloat, whose definition is used internally, but whose public API is still in the process of stabilization.

v0.2.7

• Optimize the karatsuba internals to avoid computing multiplication of zeros (sounds obvious but wasn't the case), and remove some branching.
• Very small improvement of the first "row" of schoolbook multiplication
• add more systematic benchmarking system (only on the repo ofc, this is not in the published crate)

v0.2.6

• Smarter allocations for addition on signed and unsigned big integers.
• Smarter signed subtraction, avoiding the creation of temporary
• Smarter resetting to zero to avoid losing allocation on signed integer

v0.2.5

Performance improvements for addition and subtraction in x86 assembly: looping doesn't affect carry flag, enabling one less jump instruction

v0.2.4

Fixes ARM compilation