A few possible (very partial) answers:
On 1.: 'wisdom' if it can be anything in philosophy is probably something like being able to intuit the most applicable philosophical framework to understand a given situation in its context, potentially within a societal and long-term moral framework rather than a narrow short-term self-interest way.
This might be knowing how to use a tomato in cuisine (see 3 below); developing a workable analysis of ethical use of animals for food (e.g. using Peter Singer's journey-of-life framework, rather than some humans-are-superior-beings idea); knowing when post-modernism in the social sciences needs to be topedoed a la Sokol's 'hoax'.
On 3. knowing 'tomato' is a 'fruit' is a simple ontological inference (following IS-A relations from the former to the latter), taking these terms as entities in any reasonable biological ontology. The information storage / retrieval (memory) is a side-issue. It's the ontology you need. The implication of this is most likely a theory of mind based on scientific realism, i.e. one that allows for epistemic activities as well as true (approximate) knowledge of reality, roughly described by ontologies, aka textbook knowledge of mind-independent phenomena, as best we know them - some of which these days are formalised and computable.
A post-modernist alternative approach might refuse to accept the innateness of tomato being a fruit (in the botanic sense), and say that its 'fruitness' is no more real that its 'utility in Italian cuisine-ness'.
I think on intelligence v wisdom, 'intelligence' can reasonably be understood in some objective computational sense (e.g. ability to attain better quality outputs / solutions for a goal and a given set of inputs). Many problems-to-be-solved involve an 'I' term, and/or judgements about external stimuli (is this good ice-cream?), and/or needs to generate an output in the real world (raise my hand in an auction). As soon as we say this, we are talking about an embodied mind, i.e. one whose ability to relate to the world requires awareness of a containing body, identity, boundary, and abilities to receive input signals and generate output. Creating a truly intelligent computer means (for many, including myself) creating an embodied mind with both complex computation to handle interface to the real world (the robotics & sensors part) as well as high-level abstract modules to do what we typically think of as 'intelligence' which is usually some kind of statistical inferencing.
If we could do all of this well enough that the resulting cyborg could a) successfully physically interact with the world, b) know what 'I' means, c) perform abstract computing using various representational forms, including language and d) could learn to create new knowledge about itself and the world over time, then I think we might have a machine that could theoretically be capable of 'wisdom'.