Optimized based on output from profiler on large sheets

[torerefsnes]

I ran across a performance issue with large spreadsheets, and investigated this using the YourKit profiler.

Two bottlenecks that could be addressed were identified: one is heavy use of a regexp in a function used to split a cell reference string, and the other because of unnecessary formatting of a value, where the formatting could be deferred until if/when the value was actually used.

With my set of files, total processing time went from 944 to 525 seconds with these changes included.

Note to reviewer: the change from regexp to a simplified method when splitting the cell reference works for all tests in the project, and on my testset. If I missed a format that could occur here, please let me know and I will add tests and update the code.

[monitorjbl]

Interesting, will have to confirm after merging. Regardless, deferring the cost of number parsing until they're used seems like a good idea.

[monitorjbl]

I'm not seeing the massive improvement you did with my 250k row test sheet. It is roughly 10% faster, so there's definitely an improvement but it's not nearly as marked as what you saw.

[torerefsnes]

If you’re interested, I’ll make my test set available.

[monitorjbl]

If it's not private data, that would be very helpful