I think it is an interesting proposition, but I think that most users will not find it practical, because their priority will be to enter the password quickly. For a random character string (mixed-case alphanumeric with special characters), average typing speed is a little less than 0.5 seconds/character, and it would take approximately 3.5 sec to type an 8-character password (providing for approximately 50 bits of entropy, which is sufficiently strong for a Bitwarden vault password).
With your idea, to allow for variability in reaction time, your predefined time intervals would have to be specified to the nearest second. I also believe that a delay of more than 4 seconds prior to a character would be too long, since this would result in a password entry time of more than 20 seconds, on average (if there are 8 characters). Thus, if the possible time delays before each character input time is either 1, 2, 3, or 4 seconds, then you are adding only 2 bits of entropy per character. Therefore, you could shorten the master password to 6 characters and still have around 50 bits of entropy (the minimum strength required for a Bitwarden vault password). On average, such a password would take 15 sec to enter, which is over four times the password entry time for a standard 8-character password without delays (or 2–3 times slower than entry of a 4-word passphrase without delays). For most users, this would likely be unacceptable.
Another pain point for users is that the proposed method would double the amount of information that a user must memorize about their master password.
The benefits that you have proposed are likely overstated. If this type of password entry is used by Bitwarden (or other major online services), then it is reasonable to assume that key-loggers will be designed to store the time of each character entered, along with the character value. With regards to shoulder-surfing, if this happens using video surveillance, then the timing information is readily available to the attacker.
Consider also that the user may write down the time intervals along with the password characters (in fact, it would be recommended practice to keep a written record of such information). Thus, if the master password is stolen, the time intervals may also be available to the thief.
All in all, since you requested my opinion, I would say that the proposed method creates significant inconveniences for the user, while providing very few advantages, if any.