fix unit test failures

[Patashu]

7:08 AM] Razenpok: Hey, Pata
I've ported Number-compatibility tests from C# version to check if Decimal operations produce same values as Number operations. You can check them on tests branch by running npm ci and then npm run test
[7:08 AM] Razenpok: It will run for a bit (takes several seconds on my PC)

7:08 AM] Razenpok: And there are a lot of test cases because combinatorial explosion
[7:09 AM] Razenpok: You can check all the source in tests folder
[7:09 AM] Razenpok: Currently, I observe a lot of "Unreachable code" errors and I'm not sure what to do with them
[7:09 AM] Razenpok: Either to disable these test cases or fix the bugs

[razenpok]

Current issues:

• Infinity representation in Decimal is weird, as it uses a e9e15 number. This produces incorrect results on unary operations (like exp or log10). Solution: Represent non-finite values in mantissa instead with exponent equals 0 (e.g. { m: Number.POSITIVE_INFINITY, e: 0 }). BreakInfinity.cs already does that.
• Rounding issues like Decimal.round(1009680).toNumber() is 1009679.9999999999 #52
• eq_tolerance is not good with small values. For example, new Decimal(0).eq_tolerance(5.323e-47, 1e-10): expected true, actual false
• A bunch of wrongs with sinh and tanh. E.g. try tanh with 7.23e222 or check test-report.html after running the tests for more examples
• Divide by zero is not working correctly. new Decimal(1).divide(0) :kekw:
• Same with pow 0 into anything. Try new Decimal(0).pow(1) and Math.pow(0, 1)
• Decimal.cmp has unreachable code which is actually reachable when you toss in some value combinations. For example 0, NaN
• Decimal.cmp produces incorrect values for non-zero value and NaN, try 1 and NaN

@Patashu I need your thoughts about what to do with these
I've also merged tests branch into master, so run tests from there

[Patashu]

• agreed on infinity representation
• not sure how to fix rounding in general, would appreciate your thoughts. maybe how did you fix it in BreakInifnity.cs?
• yeah should figure out what the eq_tolerance bug is. bad math or buggy implementation?
• not surprised some of the trig/hyperbolic trig functions are wrong lmao
• damn divide by zero is broken? nooo lmao
• I guess I never bothered to add a special case for pow with 0 as base? should be an easy fix
• in general I didn't proof the library very hard against nan/inf, so not surprised there are some wacky ones like that, but I think correctness is important enough that it's worth covering all cases even the weird ones

[razenpok]

Would like to pick up some of these? I'm going to fix the infinity one at least

[Patashu]

next time I'm in the mood to code I'll do whatever you haven't done yet (optimistically it'll be sometime this week)

[razenpok]

Good news: most of the broken tests were non-finite-values-related *proceeds with fixing infinity on master*

[razenpok]

I would suggest grabbing hyperbolic functions since I have no idea how they work %)
I'll probably do something with the rounding issue

[Patashu]

I got them from slabdrill iirc lol

but I can figure out a good replacement