Consistent Rot3 printing

[varunagrawal]

This PR unifies the printing functionality of Rot3 to use the Eigen::IOFormat formatting, thus ensuring consistency.

The aim would be to help the user visually identify more easily what the rotation matrices are, as well as copy them into another language such as python.

I've also

1. updated the printing of Pose3, NavState and Rot3 to provide default spacing when using the custom prefix string
2. removed extra newlines by removing newlines in the << overload
3. removed endl in << so as to avoid unnecessary buffer flushes.

For the below snippet,

Pose3 p;
cout << p << endl;
p.print();

the previous format looked like this:

# result of cout
|1, 0, 0|
|0, 1, 0|
|0, 0, 1|

[0, 0, 0]';

# result of .print()
R:
[
	1, 0, 0;
 	0, 1, 0;
 	0, 0, 1
]
[0, 0, 0]';

The new format would give the output:

# result of cout
R: [
	1, 0, 0;
	0, 1, 0;
	0, 0, 1
]
t: [0, 0, 0]'
# result of .print()
R: [
	1, 0, 0;
	0, 1, 0;
	0, 0, 1
]
t: [0, 0, 0]'

---

This change is 

[varunagrawal]

Additionally, this is how NavState now prints

R: [
	1, 0, 0;
	0, 1, 0;
	0, 0, 1
]
p: [0, 0, 0]'
v: [0, 0, 0]'

[dellaert]

Ask other people for reviews ?

[dellaert]

Ask other people for reviews ?

Dude. I don’t mean ask everyone at once :-)

[jlblancoc]

LGTM! Only, see a couple of comments above.

[varunagrawal]

@dellaert sorry about that. I figured casting a wide net would yield faster results. 😅

[varunagrawal]

@dellaert @jlblancoc now that 4.0.3 is out, I would really like to see this PR merged in. 🙂

[jlblancoc]

It might seem I'm obsessed with this tiny performance issue but... again... std::endl -> \n ? :-)

[varunagrawal]

Whoops. Yeah the other PR is based off this one hence you're seeing the issue repeat itself. Give me a few quick minutes to update this.

[varunagrawal]

So I am a bit confused about the endl to \n conversion left here. From my understanding, we want to do a buffer flush when calling print() and that is the only place where we use endl. :-)

[jlblancoc]

So I am a bit confused about the endl to \n conversion left here. From my understanding, we want to do a buffer flush when calling print() and that is the only place where we use endl. :-)

Hmm... I'm thinking of, say, we call print() on a Values or an entire FactorGraph. One flush at the end is ok... but 1000s of them, after printing (recursively) each value/matrix/vector/factor/etc. seems not necessary :-)

Still, it's probably a minor point, so feel free of leaving it as you think it's better.

[varunagrawal]

I completely agree with your point. The issue is that the Values/FactorGraph print method just calls the internal print method of the various values/factors, so we need to add the endl over there. I think that should be left to a future PR?

[jlblancoc]

Ok. I don't want to stop you doing anything more productive!
Merge it when you want to.

[varunagrawal]

Merging because Travis keeps timing out and we will be moving to Github Actions.