ts-morph incorrectly transforms nodes which begin with whitespace

if you have a node such as:

```
Foo is great
```

and you do a transform:

```
node.transform((traversal) => {
if (ts.isJsxText(traversal.currentNode) && traversal.currentNode.text != 'Foo'){
return traversal.factory.createJsxText(' is NOT great', false);
} else {
return traversal.visitChildren();
}
});
```

the result should be:

```
Foo is NOT great
```

but it is:

```
Foo is NOT great
```

instead. Notice how an extra space was added. This is due to the following error in the code:

```
function handleTransformation(oldNode, newNode) {
if (oldNode === newNode && newNode.emitNode == null)
return;
const start = oldNode.getStart(compilerSourceFile, true); //This is the error: start should be oldNode.pos which correctly accounts for whitespace
const end = oldNode.end;
let lastTransformation;
while ((lastTransformation = transformations[transformations.length - 1]) && lastTransformation.start > start)
transformations.pop();
const wrappedNode = compilerFactory.getExistingNodeFromCompilerNode(oldNode);
transformations.push({
start,
end,
compilerNode: newNode,
});
if (wrappedNode != null) {
if (oldNode.kind !== newNode.kind)
wrappedNode.forget();
else
wrappedNode.forgetDescendants();
}
}
```

As I noted in the comment, it is using an incorrect way of getting the start of the node and should be using .pos.

Locate the handleTransformation function in the ts-morph transformation code and inspect how it calculates the node start when leading whitespace is present. Update the start position handling as described, then verify that transforming the nested JSX example produces one space before "is NOT great" rather than two.

Written by the indexing model from the issue text.