Missed optimisation: lea that evaluates to same offset as sp

[ecm-pushbx]

Something I noticed in https://github.com/FDOS/kernel/pull/81#issuecomment-1146652810 it uses lea ax, [bp - 18h] just after a sub sp that computes the exact same address.

7C80:2E7E 55                push    bp
7C80:2E7F 89E5              mov     bp, sp
7C80:2E81 83EC18            sub     sp, +18
7C80:2E84 8D46E8            lea     ax, [bp-18]
7C80:2E87 50                push    ax
7C80:2E88 B81200            mov     ax, 0012
7C80:2E8B 50                push    ax
7C80:2E8C E844D3            call    01D3
-
7C80:2E8F 8B46E8            mov     ax, [bp-18]
7C80:2E92 89EC              mov     sp, bp
7C80:2E94 5D                pop     bp
7C80:2E95 C3                retn

[ecm-pushbx]

NB the optimisation allowed here is to use mov ax, sp. The choice to use push sp would not be portable as its behaviour differs depending on the CPU, so is not an alternative here.