bug(web): `teh` does not get corrected to `the` with autocorrect 🚂

[mcdurdin]

The thresholds for correction may not be quite right -- some common corrections do not seem to be applied:

• teh -> the

I will add others as I find them.

keyman for android 18.0.99-alpha

[jahorton]

I probably set the auto-correction to be a bit too cautious. We'll want to look at the following block and ease up on its thresholding:

keyman/web/src/engine/predictive-text/worker-thread/src/main/predict-helpers.ts

Lines 615 to 622 in 0ae160c

	// Find the highest probability for any correction that led to a valid prediction.
	// No need to full-on re-sort everything, though.
	const bestCorrection = suggestionDistribution.reduce((prev, current) => prev?.correction.p > current.correction.p ? prev : current, null).correction;
	if(bestCorrection.p > bestSuggestion.correction.p) {
	// Here, the best suggestion didn't come from the best correction.
	// Is it actually reasonable to auto-correct? We're probably just very
	// biased toward its frequency. (Maybe a threshold should be considered?)
	return;

What it currently does: if any te- suggestion is found, it will prevent any th- (i.e, non te-) suggestions from being auto-selected (and thus, from performing an auto-correct). As ten is close to teh - n neighbors h on US layouts - that's enough to trigger a 'block' here. (Again, currently.)