Add isunit and change norm field to property

[sethaxen]

As suggested in #60 (comment), this PR adds an exported function isunit for checking if a number is normalized. It also changes the norm field to a (deprecated) property, which just forwards to isunit. All constructors and shorthands that use take a boolean argument for whether the number is unit are also deprecated.

This is not a breaking change. However, it should wait until #74 is merged so that we are sure it doesn't break existing functionality.

This will close #60

[sethaxen]

This naming is most intuitive (we say a "unit quaternion"), but issign would be more consistent with the definition that x == sign(x).

[mikmoore]

I made a comment on #60 about the default tolerance for issign (or isunit or whatever we call it). Depending on where we use issign, I'd have different thoughts. If we leave shortcuts in inv, normalize, and other functions based on it, the tolerance should be quite strict (maybe 10eps(T)). If we don't use it internally but just have it for user convenience, we can make it a bit looser. In either case, we should have an optional argument to specify the tolerance.

Of course, I'd certainly support keeping this PR limited to a simple replacement of .isnorm and deferring changes to shortcuts to a separate PR.

[hyrodium]

I think we were planning to release breaking releases twice.

• Change norm field to property (q1.norm&q2.norm does not imply (q1*q2).norm with this change.)
• Remove norm property (q.norm doesn't work with this change.)

How about combining these changes in one breaking release?
I thought it will be okay with adding a warning to getproperty(::Quaternion, :norm), and then remove it.

[sethaxen]

How about combining these changes in one breaking release?

Yeah, makes sense to me.

I thought it will be okay with adding a warning to getproperty(::Quaternion, :norm), and then remove it.

If we remove q.norm, then I don't think we should raise a warning in this case. It's like keeping a deprecation warning after removing a deprecated function.

This PR is very stale, and I'll open a new one.

[hyrodium]

#108 will be merged soon, so I'll close this PR.