Fix dynamic rate step bracketing interval

[matias-martini]

Context

Hey there! 👋

In this PR, there was a precision issue we were having in the find_bracketing_interval function when searching for bracketing intervals at very large rates. Originally, the max rate was capped at 10.0, but we've now increased that to 1E300. The fixed step size approach wasn't cutting it, leading to floating-point precision problems as the rate got larger.

To address this, I introduced a dynamic step size that adjusts as the rate increases. This allows us to explore higher values effectively (all the way up to 1E300!) without running into precision issues. The new approach uses a logarithmic-like step, ensuring better coverage and stability across the entire range.

Changelog

• Fixed: The find_bracketing_interval function now supports extremely high rates, improving precision for very high XIRRs (over 1E300%). Dynamic step sizing ensures accurate results even at extreme rates.

QA

Now the extremely good investment converges in the specs:

  it 'calculates xirr for an extremely good and unrealistic investment' do
    cashflows = [
      [-1000, Date.new(1985, 1, 1)],
      [6000, Date.new(1985, 1, 2)]
    ]

    result = FastXirr.calculate(cashflows: cashflows)
    expect(result).to be_within(1e-6).of(1.0597571969623571e+284)
  end

[jecastro1]

LGTM