cbalance error tolerance is not uniform

[beharrop]

@wlin7 uncovered the fact that in EcosystemBalanceCheckMod.F90 the error tolerance is not the same in all subroutines. In the GridCBalanceCheck a separate error_tol parameter is still hard-coded to 1e-8_r8. To be consistent with the tolerance used by the other subroutines, it is suggested that error_tol be scrubbed out and replaced with balance_check_tolerance when checking carbon balances. Without this change, the model will crash on rare occasion with carbon balance errors -- as happened to me while running one of the SSP370 ensemble members.

[wlin7]

Thank you @beharrop for exposing and reporting the issue. This will be fixed via #7425 for master repo and #7426 for maint-3.0.