As we know in frequency domain V(s)= I(s)Z(s) then in time domain it should be V(t)= I(t) Convolution Z(t) instead of V(t)= I(t)Z(t) ?
The expression \$V(s) = I(s) Z(s)\$ is only valid if the impedance \$Z(s)\$ is time-invariant (i.e. it is not \$Z(s,t)\$). For a time-varying resistor, the time-domain expression \$v(t) = R(t) i(t)\$ is correct.
Yes, \$ V(s) = I(s) Z(s) \$ is only valid for LTI (linear, time-invariant) circuits.
For an LTI circuit, all current-voltage relations are linear expressions using linear operations (multiplying by constant, addition/subtraction, time derivative, and time integral). Therefore, conversion to frequency (Laplace) domain only involves linear equations, making analysis easy because properties of linearity can be exploited.
For general circuits, you would have to apply convolution to solve, but these cases are rare if any. Usually, you “linearize” a nonlinear system by analyzing it under specific conditions/assumptions where linear approximations are valid. An example is small-signal AC analysis for transistors.