Where do I start to study counterpoint from a melodic (ie, melodies
interaction) point of view without harmony?
In the study of 16th Century counterpoint, there is the notion of what properties normally characterize 'good' melodies. The standard text for this is "Gradus ad Parnassum" or "The Study of Counterpoint" by J. J. Fux.
For example, good melodies usually have a highest or lowest note that serves as a climax, and generally should not be repeated. Normally there should be no more than two skips in the same direction, etc. Most of these rules are for the melody around which other voices will harmonize with and the examples are usually in whole notes, and melodies involving mixed durations have more complicated rules.
However, I am wondering if your goal of studying counterpoint makes sense since the primary purpose of counterpoint during the common practice period was to write two or more melodies that harmonize with each other. That includes dissonances that are resolved. Those harmonies are chords (I am including two note intervals as chords).
So in the end, it has been said that good harmony and counterpoint are two different ways at approaching music. In harmony, you choose the chord progressions with the voice leading a secondary consideration. In counterpoint, the primary focus is the voice leading, and the harmony is secondary.
Can modal counterpoint be studied without studying harmony?
Currently, students normally study harmony first. But, historically, counterpoint preceded harmony, and in fact our understanding of harmony began with composers discovering which intervals sounded good when writing two or more simultaneous melodies.
One last point regarding the Fux text (which can be found online for free at IMSLP), is that it discusses cadences in each of the modes. If you're applying this to Classical Indian music, I'm not sure how valuable that will be though.