Be cautious about underfollow

[dantaik]

In proveIncorrectLookahead, I'm uncertain whether _lookaheadTail -= 1 could underflow under normal conditions. Test coverage might already be sufficient, though I haven't reviewed the test code yet.

 // If it is the current epoch's lookahead being proved incorrect then insert a fallback preconfer
        if (block.timestamp < epochEndTimestamp) {
            uint256 _lookaheadTail = lookaheadTail;

            uint256 lastSlotTimestamp = epochEndTimestamp - PreconfConstants.SECONDS_IN_SLOT;

            // If the lookahead for next epoch is available
            if (lookahead[_lookaheadTail % LOOKAHEAD_BUFFER_SIZE].timestamp >= epochEndTimestamp) {
                // Get to the entry in the next epoch that connects to a slot in the current epoch
                while (lookahead[_lookaheadTail % LOOKAHEAD_BUFFER_SIZE].prevTimestamp >= epochEndTimestamp) {
                    _lookaheadTail -= 1;
                }

                // Switch the connection to the last slot of the current epoch
                lookahead[_lookaheadTail % LOOKAHEAD_BUFFER_SIZE].prevTimestamp = uint40(lastSlotTimestamp);

                // Head to the last entry in current epoch
                _lookaheadTail -= 1;
            }

            lookahead[_lookaheadTail % LOOKAHEAD_BUFFER_SIZE] = LookaheadBufferEntry({
                isFallback: true,
                timestamp: uint40(lastSlotTimestamp),
                prevTimestamp: uint40(epochTimestamp - PreconfConstants.SECONDS_IN_SLOT),
                preconfer: getFallbackPreconfer(epochTimestamp)
            });

            _lookaheadTail -= 1;

            // Nullify the rest of the lookahead entries for this epoch
            while (lookahead[_lookaheadTail % LOOKAHEAD_BUFFER_SIZE].timestamp >= epochTimestamp) {
                lookahead[_lookaheadTail % LOOKAHEAD_BUFFER_SIZE] =
                    LookaheadBufferEntry({isFallback: false, timestamp: 0, prevTimestamp: 0, preconfer: address(0)});
                _lookaheadTail -= 1;
            }
        }

[AnshuJalan]

The original pre-test code did have an underflow issue. It was detected during tests and duly fixed.