(a) answer: The difference comes from how the two reagents are bonded. KCN is a mostly ionic compound, so in it the cyanide exists as a free $\mathrm{CN^-}$ ion. In this ion the carbon end is the better (more nucleophilic) site, so carbon attacks the haloalkane and we form a C to C bond, giving an alkyl cyanide, $\mathrm{R\text{-}CN}$. AgCN, on the other hand, is mostly covalent, so the carbon lone pair is tied up with silver and is not free; only the nitrogen lone pair is available to attack. So nitrogen bonds to the carbon of the haloalkane, giving an isocyanide, $\mathrm{R\text{-}NC}$.
(b) answer: $\mathrm{S_N2}$ needs the nucleophile to attack the carbon from the back side, so it works best when that carbon is easy to reach and when the C to Cl bond is not too strong. In benzyl chloride the chlorine sits on an $\mathrm{sp^3}$ benzylic carbon, which is open to backside attack, and the bond is a normal single bond, so $\mathrm{S_N2}$ happens easily. In chlorobenzene the chlorine sits directly on an $\mathrm{sp^2}$ aromatic carbon. Here the lone pairs of chlorine go into resonance with the ring, giving the C to Cl bond partial double-bond character, which makes it short and strong, and the flat ring also blocks backside attack. So chlorobenzene is far less reactive towards $\mathrm{S_N2}$ than benzyl chloride.
(c) answer: A Grignard reagent, $\mathrm{RMgX}$, has a very polar carbon to magnesium bond where the carbon is strongly negative, almost like a carbanion. Such a carbon grabs any acidic hydrogen very eagerly, and water has exactly such a hydrogen. So if even a trace of moisture is present, the Grignard reagent is destroyed before it can do its intended job, turning into an alkane:
\[ RMgX + H_2O \rightarrow RH + Mg(OH)X \]
That is why the reagent must be prepared under dry (anhydrous) conditions.