Improve validation in the cutting reconstruction function

[garrison]

• I took the first few validations and moved them into the following if block, since the conditionals are a bit redundant. This looked better to me once I wrote everything out.
• The dict can really be any Mapping, so I use that broader class when calling isinstance. At the same time, I've left dict in the type annotations for the sake of users who are reading the docs but don't know what a "Mapping" is.
• The thing that prompted this was the last chunk, which validates the number of subexperiments. The lack of this error led to an experience where @ibrahim-shehzad was getting "wrong" results. (It turned out, the observables were different than the ones used when generating the subexperiments; having this check will at least make sure the number of observables makes sense.)

Remaining action items

• Add tests to bring coverage to 100%

[coveralls]

Pull Request Test Coverage Report for Build 9104575058

Details

• 17 of 17 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.01%) to 95.484%

---

Totals
Change from base Build 9104465513:	0.01%
Covered Lines:	3510
Relevant Lines:	3676

---

💛 - Coveralls

[caleb-johnson]

This came about because utility-scale workloads may be run in chunks, saved to disk, and recombined, which means there could very easily be errors during that recombination?

[garrison]

That is indeed one motivation. Mostly I just wanted to add this as an additional sanity check when passing over this code, since the below lines implicitly make this assumption.

[caleb-johnson]

Looks good, thanks! 👍