Allow BigDecimal accept Float without precision

[mrzasa]

Make sure that conversion from a Float works consistently with the conversion from a string representing that float.

Current master state:

pry(main)> BigDecimal(1.2)
ArgumentError: can't omit precision for a Float.
from (pry):22:in `BigDecimal'
pry(main)> BigDecimal('1.2')
=> 0.12e1

After this changes:

pry(main)> BigDecimal(1.2)
=> 0.12e1
pry(main)> BigDecimal('1.2')
=> 0.12e1

When precision is omitted it's set to 0.

Fixes #213
Blocked by #313

[mrkn]

@mrzasa Could you add more assertions to check many cases, especially for numbers that don't have accurate representations in Float?
Moreover, we must compare two values by assert_in_delta and pass an appropriate value to the delta argument for the original Float value; it should be 0.5ulp.

[mrkn]

I reconsidered how to check here.

I suggest comparing the result of to_f with the original Float value, like this: assert_equal(4.2, BigDecimal(4.2).to_f).
This approach is preferable because it doesn’t depend on the specific decimal representation for a given Float value. I’d like to avoid relying on the particular algorithm used for converting a Float to a BigDecimal.