Performance optimizations

[Cadienvan]

Did some small, yet significant performance improvements, explained below:

• Using a Buffer.concat and generally reducing the amount of operations made by the baseId function.
• Used a bitwise operator for the generate -> count variable, as they are generally faster than arithmetic ones.
• Removed the pad function as its only usage can be converted, as did, to a sliced string.
• As the '==' padding used in the baseId function is always the same, I brought it out of the function scope preventing a redundant operation being done every time.

All tests seem to pass and benchmarks provide a +1/+5% benefit.

Before updates, on a Macbook Pro 2021 (M1 Max, 32GB RAM):

hyperid - variable length x 14,254,472 ops/sec ±0.62% (88 runs sampled)
hyperid - fixed length x 14,098,171 ops/sec ±0.69% (90 runs sampled)
hyperid - fixed length, url safe x 14,382,843 ops/sec ±0.75% (91 runs sampled)

After updates:

hyperid - variable length x 14,481,182 ops/sec ±0.52% (84 runs sampled)
hyperid - fixed length x 14,268,481 ops/sec ±0.57% (87 runs sampled)
hyperid - fixed length, url safe x 14,797,637 ops/sec ±0.62% (92 runs sampled)

I reached 15M ops/sec in a couple of benchmarks, yet those results are the medians.

[mcollina]

lgtm, good work

[Cadienvan]

Hi @mcollina , just to be sure I didn't miss anything, PR is still open, you still waiting something from me or just didn't have time to pull in?

[mcollina]

Shipped v3.1.0