Ultimately, this is an impossible question. Lich provides fantastic defensive bonuses, but a sufficiently-optimized wizard would be able to do better without giving up spellcasting. So at high optimization, lich is not worth any LA, but is clearly too good to be free.
A better approach that I have used is 4e-style epic destinies. Basically, in 4e, at the end of the game (21st through 30th; 4e goes to 30th level), every character chooses an epic destiny, which pretty much always ends with a feature that gives partial/conditional immortality, as lich does—and indeed, lich is one of the options. I have introduced a similar idea to my 3.5 games to some success, and it's a great way to implement lich. The trick is to come up with similar-but-distinct forms of immortality for other types of characters. 4e can provide a lot of inspiration for this; you don’t even really need the rules, just concepts.
Wizards did back-port epic destinies to 3.5 in this web article, though they implement them as replacing epic feats between 21st and 30th—arguably truer to 4e, but that means getting into 3.5 epic rules, which are bad. A variant at the bottom has them replacing feats before 20th, but personally I just do it as a pure add-on thing, without replacing those feats. As long as everyone has one, balance is OK.
More problematically, lich isn't one of those that Wizards back-ported. For that, look to the dread necromancer class from Heroes of Horror, which gradually becomes more lich-like before gaining the template at 20th as a capstone feature. Selecting some of those class features for 12th, 15th, and 18th works well.