Optimize transition lookup in CompiledAutomaton.addTail using binary search

### Description

Description:

### Description
In `CompiledAutomaton.java` within the `addTail` method, a linear scan $O(N)$ is currently used to find the largest transition index where the transition's minimum label is less than the `leadLabel`:
```java
// Find biggest transition that's < label
// TODO: use binary search here
int maxIndex = -1;
int numTransitions = automaton.initTransition(state, transition);
for (int i = 0; i < numTransitions; i++) {
automaton.getNextTransition(transition);
if (transition.min < leadLabel) {
maxIndex = i;
} else {
// Transitions are always sorted
break;
}
}
```

Since the automaton transitions are always sorted by minimum label first, we can optimize this transition lookup to run in $O(\log N)$ time by implementing a binary search using random access (`automaton.getTransition(state, index, transition)`).

Start in CompiledAutomaton.java at the addTail method and inspect how automaton.getTransition(state, index, transition) provides random access. Replace the sorted-transition scan with a binary search for the largest minimum label below leadLabel, preserving the existing maxIndex behavior. Done means the lookup is logarithmic while selecting the same transition boundary.

Written by the indexing model from the issue text.