Use input structure for get_number_of_species

[pmrv]

This fixes an issue where tables are getting slowed down a lot for jobs run before #1341 was merged. For those jobs output/structure/species is missing and accidentally trigger decompression of the jobs.

[coveralls]

Pull Request Test Coverage Report for Build 8117105299

Details

• 1 of 1 (100.0%) changed or added relevant line in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 93.174%

---

Totals
Change from base Build 8108226849:	0.0%
Covered Lines:	14209
Relevant Lines:	15250

---

💛 - Coveralls

[jan-janssen]

Initially I thought this breaks the compatibility with the SQSJob but it stores its structures in output/structures so it is not effected. Still this made me wonder if it would be better to search for a structure in multiple places, start with output/structure then maybe output/structures and finally input/structure. Another case where this breaks the functionality is a LAMMPS calculation with variational grand canonical sampling.

[pmrv]

Initially I thought this breaks the compatibility with the SQSJob but it stores its structures in output/structures so it is not effected. Still this made me wonder if it would be better to search for a structure in multiple places, start with output/structure then maybe output/structures and finally input/structure.

I can do that. Personally I like input/, because that's somewhat consistent with the other predefined functions.

Another case where this breaks the functionality is a LAMMPS calculation with variational grand canonical sampling.

In a round about way actually no, but only because in case of SGC MC a lot of our code breaks when not all elements are present in the initial structure.