They system you are suggesting would in effect turn every trial into a little election on what the law should be, but with only a small fraction of the electorate voting, and the results only applying to that one case.
And also, every case adds to the case history; it doesn't take many cases to create a pattern. So no, the results don't only apply to that one case. And even if they do, a single instance of well-applied nullification might not mean much to the general public, but it means the world to the defendant.
The legislature does no such thing. Theoretically, representatives do their job by pandering to constituent majorities.
Say someone is on trial for tax evasion, because they paid only 20% rather than the 30% they owe under the law. Is it OK for a juror to say "gee, taxes should only be 20% anyway, so I'm going to let him off the hook"?
On the other hand, with a tax rate of 90% (to use an extreme example), juries might start consistently acquitting people for not paying their taxes, and that sends a message.
In any case, I don't want to see any measures taken to stop the scenario you describe. It'll work out just fine on its own.
And this doesn't only apply to tax laws. There could be a difference of opinion on other cases too.
While the law nullification idea is a good idea in theory, it starts to break down in practice.
Of course it's better to make a system that would be consistently fair (e.g. keep taxes at the level vast majority of society supports it). But if two available options are "not fair" and "not always fair" - I'd pick "not always fair".
If that means we get a few more "not guilty" verdicts than we should, so be it. We make the same tradeoff with "innocent until proven guilty" and "beyond a reasonable doubt": better to let guilty people free than convict innocent people. The difference: we need "beyond a reasonable doubt" applied to laws, not just to people.