Serious calculation bug

[DHMike57]

I was having a look at issue #889 - on opening test7.sc and and manually making the test change A7=90 the actual results calculated are wrong. For example G20 becomes 955958 instead of 3178634 (checked by a couple of other spreadsheets including vanilla sc). The behaviour varies with the number of rows and columns and with whether or not autocalc is set.

Investigation with git bisect shows that the regression was introduced by commit e9971440f21f52f5d15c0a9e9c694d790a98fba3 in May 2021.

PR #907