eckey: speed up _pubkey_tweak_add by using ecmult_gen_var

Rather than using the generic double multiply (with na=1), use the
newly introduced fast variable-time generator point multiplication
and add it up to the base public key manually.

On arm64, this improves the performance of the
`secp256k1_ec_pubkey_tweak_add` API function by about 80%:

----- Before (prior this commit): -----
```
$ ./build/bin/bench tweak
Benchmark                     ,    Min(us)    ,    Avg(us)    ,    Max(us)

ec_pk_tweak_add               ,    16.1       ,    16.2       ,    16.6
```

----- After (this commit): -----
```
$ ./build/bin/bench tweak
Benchmark                     ,    Min(us)    ,    Avg(us)    ,    Max(us)

ec_pk_tweak_add               ,     8.94      ,     8.98      ,     9.29
```

Note that the following API functions also benefit from the improved code path:
- secp256k1_xonly_pubkey_tweak_add
- secp256k1_xonly_pubkey_tweak_add_check
  (this one is consensus-critical for P2TR script path spends, see BIP341)
- secp256k1_keypair_xonly_tweak_add
- secp256k1_musig_pubkey_ec_tweak_add
- secp256k1_musig_pubkey_xonly_tweak_add

Author: theStack

src/eckey_impl.h

@@ -61,8 +61,8 @@ static int secp256k1_eckey_privkey_tweak_add(secp256k1_scalar *key, const secp25
 
 static int secp256k1_eckey_pubkey_tweak_add(secp256k1_ge *key, const secp256k1_scalar *tweak) {
     secp256k1_gej pt;
-    secp256k1_gej_set_ge(&pt, key);
-    secp256k1_ecmult(&pt, &pt, &secp256k1_scalar_one, tweak);
+    secp256k1_ecmult_gen_var(&pt, tweak);
+    secp256k1_gej_add_ge_var(&pt, &pt, key, NULL);
 
     if (secp256k1_gej_is_infinity(&pt)) {
         return 0;