Flatten(): render even large numbers as-is, not using scientific notation

[Al2Klimov]

E.g. 2000000000000000000 (explicitly), not 2e+18 (as with fmt.Sprintf("%v")).

fixes #659

TODO

• review (not @Al2Klimov)
• unit test the whole function as already on it (@Al2Klimov)
• transfer Do not convert/calculate on custom variable content #659
• adjust "fixes Do not convert/calculate on custom variable content #659"

[julianbrost]

Anyway, the tests don't have to wait for the implementation details. Part of the tests should be to verify that the implementation isn't changed in other ways, so please add them as a first commit so that they pass on the old implementation and then add another case for large floats with the updated implementation.

[julianbrost]

Please reorder the commits so that that the tests are added first without the 1e23 case, so that when the tests are run on this commit, it shows that the old implementation conforms to the tests, and then fixing the implementation including an additional test case, showing that the new implementation passes the same tests as the old one plus a new one.