My answer will be divided into 1) a short discussion of rating manipulation because you explicitly mention farming, farmbitrage, rating manipulation, and your liking of rating gains, 2) whether it is wrong to challenge only weaker opponents and 3) a short conclusion. The first point is also relevant, because in deontological ethics the intention / the characteristic of the action itself is the relevant factor for judging the morality of an action, whereas in consequentialist ethics, the outcome of an action is the basis for the assessment of the morality of an action.
1) Rating manipulation
In my view, it is not right to manipulate one's rating (no matter why or whether it's rating inflation or sandbagging). The rating aims to have a fair and accurate strength estimation of the players. Players use these ratings to get interesting and fairly matched games.
The main issue is that you're impacting other players if successfully inflating your rating.
If your rating is not accurate, the rating adjustments of the affected players will also not be accurate. As you stated that you love winning, I think you can relate to people not wanting their ratings to be too high, as they will be matched with stronger opponents and therefore are less likely to win games.
Other people like to play strong opponents in order to have challenging games, and if you are rating-matched in a game, your opponent may not have a game that is as interesting as he'd like it to be.
Similar arguments can be made if the rating gets artificially decreased ("sandbagging", which in my opinion is even worse than rating inflation).
Another issue is dishonesty, which is also generally morally wrong (for most people, in deontological ethics). However, from a consequentialist viewpoint, it may not be that bad - provided you enjoy the numbers only for yourself and do not brag about them or use them to pretend false facts to others. If no one gets harmed, then it is OK from a consequentialist point of view.
2) Playing weaker opponents only
If you are playing enough games and play normally (without any assistance), then it is absolutely OK to play with weaker opponents (from the consequentialist point of view).
This is due to the rating mechanism: It is hard to inflate one's rating. Sandbagging is easy. But for lasting rating boosts, you would most likely need 1) to stop playing after reaching a high rating (and probably keep a provisional rating), or 2) to cheat in order to maintain it (I do not need to tell you that this would be another immoral thing to do).
That is because your rating will approach your "true" rating the more games are played. For each game result, your rating change will depend on the expected score, and if you for example lose against a much lower-rated opponent who is close to your actual playing strength, the rating loss will be high; but if you win the rating gain will be low. This is why the rating mechanism works and everyone is expected to get an appropriate rating with enough played games.
This means, that the rating you can gain by playing weaker opponents is capped and therefore it does not matter whether you play strong or weak opponents - your rating adjustments take this into account.
3) Conclusion
If applying deontological ethics, it could be considered problematic (depending on the concrete moral values, e.g. regarding dishonesty or manipulation). In case your intention of "farming" weaker players is to increase your rating above the correct level, then this action can be considered wrong even if not successful! The attempt of doing something immoral cannot be moral itself (for example, it would violate Kant's categorical imperative).
If applying a consequentialist ethic, it is not unethical to (try to) gain rating by playing against only weaker players (farming) because the rating mechanism prevents adverse effects on others (i.e. a successful rating manipulation).
However, if you were successfully inflating your rating the arguments from section 1) would also apply in the consequentialist framework. One way to make it more acceptable in that unlikely case would be to play in a "clique" of players who only play each other - and no one outside, so that the ratings, which may be meaningless or inaccurate do not affect other players from the public pool.
If you ask me personally, I share the latter point point of view - if it makes you happy, play weaker opponents.