Handle items with >10⁶ elements in Excel data I/O

[khaeru]

In #289 (comment), @zikolach pointed out that some Scenarios may have parameters with more than 1 million elements, that might exceed the size limit of the Microsoft Excel format or the Python tools used to read/write it.

Scenario.{read,write}_excel() and the associated format should be:

• Tested for their behaviour with such 'large' data, and
• Extended to support round-trip write and read of such data.

[khaeru]

Further suggestion from @zikolach at #289 (review):

In general would be good to allow splitting items data into multiple sheets as we can reach rows limit. We could avoid to have separate sheet with item name-type mapping by embedding type into sheet name, e.g. type|name|n or similar convention, where type is item type, name - item name and n is a number of sheet for "large" items.

That sounds like a good proposal.

Users may have old Excel spreadsheets sitting around and will expect them to continue working. So we would need to add a second code path for these alternate formats; support both for some time; and then deprecate & eventually remove the current one. We would need an explicit decision about when the current format will no longer be supported.