StepTracker (both notebooks): top-2/top-3 frequencies duplicated + real-time peak slope wraps at buffer start

[jonfroehlich]

Found in the #16 correctness/prose/comments audit (verified by running on the default log). Both bugs are present in both StepTracker notebooks (StepTracker-Exercises.ipynb and StepTracker-WithExampleSolution.ipynb).

1. Top-2 / top-3 frequencies are copy-paste duplicates. The "top 3 frequencies" extraction assigns all three to the same element:

top1_freq = freqs[ind_sorted_by_coef][0]
top2_freq = freqs[ind_sorted_by_coef][0]   # should be [1]
top3_freq = freqs[ind_sorted_by_coef][0]   # should be [2]

Verified on the default file the sorted top-3 are [1.545, 1.636, 3.091] Hz, but all three vars get 1.545. The variables aren't reused downstream (so no output changes), but the code is wrong and misleading to students.

2. Real-time peak detector wraps at the buffer start. In the per-buffer peak loop:

for j in range(0, len(buffer) - 1):
    forward_slope  = buffer[j + 1] - buffer[j]
    backward_slope = buffer[j] - buffer[j - 1]   # j==0 -> buffer[-1] wraps to the LAST sample
    if forward_slope < 0 and backward_slope > 0:
        ...

At j == 0, buffer[j-1] is buffer[-1] (the last sample), so the slope test compares across the buffer seam and can flag a spurious peak at index 0. Masked by the min_peak_height threshold for the stock data, but it's a genuine off-by-one in a stub students extend. Fix: iterate range(1, len(buffer) - 1) (the one-sample buffer overlap already covers the seam).